Tilburg University
Psychological sentiments and economic behavior
van de Ven, J.
Publication date:
2003
Link to publication
Citation for published version (APA):
van de Ven, J. (2003). Psychological sentiments and economic behavior Tilburg: CentER, Center for Economic
Research
General rights
Copyright and moral rights for the publications made accessible in the public portal are retained by the authors and/or other copyright owners
and it is a condition of accessing publications that users recognise and abide by the legal requirements associated with these rights.
- Users may download and print one copy of any publication from the public portal for the purpose of private study or research
- You may not further distribute the material or use it for any profit-making activity or commercial gain
- You may freely distribute the URL identifying the publication in the public portal
Take down policy
If you believe that this document breaches copyright, please contact us providing details, and we will remove access to the work immediately
and investigate your claim.
Download date: 28. jul. 2017
Psychological Sentiments
&
Economic Behaviour
Psychological Sentiments and Economic Behaviour
PROEFSCHRIFT
Ter verkrijging van de graad van doctor aan de
Universiteit van Tilburg, op gezag van de rector
magnificus, prof.dr. F.A. van der Duyn Schouten, in het
openbaar te verdedigen ten overstaan van een door het
college voor promoties aangewezen commissie in de aula
van de Universiteit op
vrijdag 10 oktober 2003 om 14.15 uur
door
Jeroen van de Ven,
geboren op 10 april 1975 te Huissen.
PROMOTOR:
Prof.Dr. Th.C.M.J. van de Klundert
To my parents,
Contents
1 Introduction
1.1 Rational economic man . . . . . .
1.2 Defending rational economic man
1.3 Economics and psychology . . . .
1.3.1 The economic paradigm .
1.3.2 Preferences . . . . . . . .
1.3.3 Beliefs . . . . . . . . . . .
1.3.4 Discounting . . . . . . . .
1.3.5 Maximization . . . . . . .
1.4 Applications . . . . . . . . . . . .
1.5 Discussion . . . . . . . . . . . . .
1.6 Overview of the thesis . . . . . .
1.6.1 Main themes . . . . . . .
1.6.2 Detailed overview . . . . .
2 The
2.1
2.2
2.3
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Economics of the Gift
Introduction . . . . . . . . . . . . . .
The Gift . . . . . . . . . . . . . . .
Motivations to give . . . . . . . . . .
2.3.1 Gifts as exchange mechanism
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
1
1
2
3
4
4
8
10
12
13
16
19
19
21
.
.
.
.
27
27
29
32
32
Contents
iv
2.3.2 Altruism . . . . . . . . . . . .
2.3.3 Fairness . . . . . . . . . . . .
2.3.4 Warm glow & social approval
2.3.5 A gift as signalling device . .
2.4 Discussion . . . . . . . . . . . . . . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
3 The Demand for Social Approval as a Motivation to Give
3.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . .
3.2 Social approval, status, and gift-Giving . . . . . . . . . . . .
3.2.1 The basic model . . . . . . . . . . . . . . . . . . . .
3.2.2 Sequential move equilibrium . . . . . . . . . . . . . .
3.2.3 Simultaneous move equilibrium . . . . . . . . . . . .
3.2.4 Gifts and markets . . . . . . . . . . . . . . . . . . .
3.3 Related literature . . . . . . . . . . . . . . . . . . . . . . . .
3.4 Concluding remarks . . . . . . . . . . . . . . . . . . . . . . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
4 On the Viability of Gift-exchange in a Market Environment
4.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
4.2 Exchange mechanisms . . . . . . . . . . . . . . . . . . . . . . .
4.2.1 Substantive and symbolic utility . . . . . . . . . . . . . .
4.2.2 Reciprocal exchange . . . . . . . . . . . . . . . . . . . .
4.2.3 Market exchange . . . . . . . . . . . . . . . . . . . . . .
4.3 Equilibrium . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
4.3.1 Complete commodification . . . . . . . . . . . . . . . . .
4.3.2 Partial commodification . . . . . . . . . . . . . . . . . .
4.3.3 Valuation, patience, and viability. . . . . . . . . . . . . .
4.4 Welfare . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
4.5 Concluding remarks . . . . . . . . . . . . . . . . . . . . . . . . .
4.6 Appendix . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
5 Welfare Gains of Labeling with
5.1 Introduction . . . . . . . . . .
5.2 Background . . . . . . . . . .
5.2.1 General background .
5.2.2 Labeling . . . . . . . .
5.3 Description of the model . . .
Heterogeneous
. . . . . . . . . .
. . . . . . . . . .
. . . . . . . . . .
. . . . . . . . . .
. . . . . . . . . .
Consumers
. . . . . . . .
. . . . . . . .
. . . . . . . .
. . . . . . . .
. . . . . . . .
.
.
.
.
.
.
.
.
.
.
34
37
40
43
51
.
.
.
.
.
.
.
.
55
56
58
58
64
71
72
76
78
.
.
.
.
.
.
.
.
.
.
.
.
81
81
84
85
88
89
91
91
93
94
96
98
101
.
.
.
.
.
105
105
107
107
109
110
Contents
v
5.3.1 Imperfect information
5.3.2 Standards . . . . . . .
5.3.3 Labeling . . . . . . . .
5.4 Standards or labels? . . . . .
5.5 Discussion and conclusions . .
5.6 Appendix . . . . . . . . . . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
6 Rewards, Self-confidence, and Motivation:
of Rewards
6.1 Introduction . . . . . . . . . . . . . . . . .
6.2 The model . . . . . . . . . . . . . . . . . .
6.2.1 Preliminaries . . . . . . . . . . . .
6.2.2 The main assumptions . . . . . . .
6.2.3 Timing and summary of the game .
6.3 Equilibrium behavior . . . . . . . . . . . .
6.3.1 A pooling equilibrium . . . . . . .
6.3.2 A semi-separating equilibrium . . .
6.4 Rewards, self-confidence, and motivation .
6.5 Discussion . . . . . . . . . . . . . . . . . .
6.6 The hidden costs of rewards . . . . . . . .
6.7 Conclusions . . . . . . . . . . . . . . . . .
6.8 Appendix . . . . . . . . . . . . . . . . . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
111
112
114
118
120
124
The Hidden Rewards
129
. . . . . . . . . . . . . 129
. . . . . . . . . . . . . 132
. . . . . . . . . . . . . 132
. . . . . . . . . . . . . 132
. . . . . . . . . . . . . 136
. . . . . . . . . . . . . 137
. . . . . . . . . . . . . 140
. . . . . . . . . . . . . 141
. . . . . . . . . . . . . 143
. . . . . . . . . . . . . 144
. . . . . . . . . . . . . 146
. . . . . . . . . . . . . 147
. . . . . . . . . . . . . 149
7 Optimal Subsidies with Rationalizing Agents:
but Don’t Subsidize Too Much
7.1 Introduction . . . . . . . . . . . . . . . . . . .
7.2 Preference management . . . . . . . . . . . .
7.3 Evidence from psychology . . . . . . . . . . .
7.4 A simple model . . . . . . . . . . . . . . . . .
7.4.1 Preferences and choice . . . . . . . . .
7.4.2 Rationalizing choice . . . . . . . . . . .
7.4.3 The effect of a subsidy . . . . . . . . .
7.5 Discussion . . . . . . . . . . . . . . . . . . . .
7.6 Conclusions . . . . . . . . . . . . . . . . . . .
8 Summary in Dutch
.
.
.
.
.
.
Subsidize Enough
157
. . . . . . . . . . . 157
. . . . . . . . . . . 160
. . . . . . . . . . . 161
. . . . . . . . . . . 164
. . . . . . . . . . . 164
. . . . . . . . . . . 169
. . . . . . . . . . . 170
. . . . . . . . . . . 173
. . . . . . . . . . . 176
179
Contents
References
vi
187
Acknowledgements
During the past four years or so, I have been working on this thesis. This has been
a very enjoyable time for me, if not only because of the topic. How amusing is it to
read about the irrationalities of other people! The negative side of studying some
of the psychological literature, however, is that you also learn a bit too much about
yourself. I realized only too well how well I fit the evidence of irrational behaviour
myself. Thus, I tend to attribute failures to other people or to forces beyond my
power, taking credit for successes myself. After a purchase, I tend to experience an
unpleasant feeling of doubt, after which I try to rationalize my choice. I certainly
don’t treat sunk costs as sunk. And worst of all, I have postponed writing this
preface until the very last moment, not being able to recall all the brilliant ideas
I had for it. All this is, to the very least, a bit disturbing: being irrational is one
thing, but being aware of it another. Some people may want to stay ignorant
about such matters, and I can only encourage them not to continue reading.
However difficult it is to give credit to others, I will have to admit that I have
been surrounded with such nice, skilful, and helpful people that any errors can
only be blamed on me. Above all, I am much indebted to my supervisor professor
Theo van de Klundert. Starting during my undergraduate studies, he has been a
continuous source of inspiration for me. I respect his broad views on economics
and his never-ending enthusiasm. As a supervisor, he has shaped many of my
current views. I will be very proud if somebody recognizes me as being one of
his students. Part of my time as a Ph.D. I spent at GREMAQ in Toulouse. It
Contents
viii
was a very stimulating environment there. Special thanks go to professor Jean
Tirole for supervising me during that time, and for providing me with ideas for a
chapter. Furthermore, I would like to express my gratitude to Eric van Damme
for encouraging and helping me to go there.
Thanks also to the committee: Professors A.L. Bovenberg, W. Güth, M.J.
James, G.J. Mellenbergh, C. van Ewijk, and Doctor J.A. Smulders. I appreciate
their time and efforts a lot and I am very proud to have them in my committee.
I was also extremely fortunate to share the office with Richard Nahuis and
Luuk van Kempen. Luuk kindly saved my ideas from the paper shredder and
we cried together on more than one occasion (having hot Argentenean soup or
vlammetjes at the drunken horse). We even managed to keep a fish in our office,
though at the sacrifice of our coffee machine. I suspect that our jogging outfit is
legendary. Nobody encouraged me more than Richard. Needless to say that I am
very happy to work at the same place as him again.
Next, I would like to thank two co-authors. It is a pleasure and honour to
work together with Enrico Diecidue and Anton Souvorov. I am impressed by
their brightness and kindness. I am confident in an enjoyable and fruitful future
cooperation.
There were many other people who helped me out on more than one occasion.
Among others, Jan Boone, Jeffrey James, Karim Sadrieh, and Sjak Smulders
always had their door open for questions and chats. I am happy that they were
so quick in understanding and fixing my problems, otherwise this preface would
probably have been postponed even longer. In addition, I would like to thank the
people who read and commented on several chapters. Among others, Martin van
Tuijl deserves credit for stimulating my interest in science.
Everyday had three focal points. With Swiss preciseness the coffee group gathered for lunch and coffee breaks (for a composition of the group, see e.g. CentER
Ph.D. thesis 111). One person in particular deserves special mention. Theo Leers
was one of the finest sorts of people who made life beautiful. He was a great young
scientist and very funny person. I just wish he would still be around.
During the past four years many colleagues became dear friends. I would like
to thank them all, in particular Riccardo.
Writing a thesis cannot be done without the help of secretaries. I would like
to thank the secretaries at the department of economics and CentER. Aude
ix
Contents
Schloeßing in particular I wish to thank. Without her help, I would have been
lost in France for sure.
My family provided continuous support throughout my life, if only indirect.
Heel erg bedankt papa, mama, oma, Margreet, Maurice, Suzanne en Thomas.
Finally, Michèle, thank you for being with me. You unite all of the above
mentioned qualities.
1
Introduction
People do not always behave as economists expect them to do. Not so long ago,
my girlfriend, who holds a Ph.D. in economics, had to decide which health club
to go to. One of these clubs had a higher membership fee than the other, but also
provided more facilities. In the end she made the decision to go for the expensive
one. Surprisingly, it was not the high price-quality ratio that ultimately mattered
most, but, she reasoned, the fact that if she paid a high contribution, she would
feel committed to actually go.
1.1 Rational economic man
Economics students are taught early in their study about the rational economic
man1 . The economic man at least knows his preference ordering (which satisfies
transitivity and completeness), and given this ordering plus some constraints he
attempts to attain his most desired bundle of goods. Moreover, economic man
is incredibly good at solving optimization programs to calculate the best, say,
consumption to savings ratio. In other words, he is usually depicted as selfish as
well as smart.
1 Rationality
is a delicate concept. For current purposes, I need not define it in a precise way. Hereafter I
drop the term and just speak of economic man and the economic paradigm (see section 1.3.1).
Chapter 1. Introduction
2
Being smart means, among other things, that decisions are based on weighing
marginal benefits against marginal costs (a consequence of maximizing utility).
This means that the optimal frequency of visiting a health club is independent
on membership fees, since such fees are fixed costs and do not influence marginal
costs. If people fail to disregard sunk costs (or do not feel that they should be
treated as such), then there is something wrong with modeling those people as
economic men. If even economists cannot be modeled as economic men, there is
something seriously wrong. In this thesis, individuals are taken to be less selfish
and smart.
1.2 Defending rational economic man
The economic man is quite often practical for reasons of tractability. Clearly,
however, it is not a very accurate description of most people in real life. Friedman
and Savage [1948] have nevertheless defended the use of economic man on the
grounds that it does not matter so much whether the assumptions underlying the
model are truly accurate, as long as the predictions are (see also Thaler [1980]).
Economic man need not literally and consciously make the necessary calculations.
In a well known passage, they compare the assumptions behind the economic man
who calculates, say, the optimal lifetime savings plan, to a billiard player who has
to predict all the movements of the balls and therefore essentially needs to solve
a system of equations. They reason:
”... it seems not at all unreasonable that excellent predictions would
be yielded by the hypothesis that the billiard player made his shots as
if he knew the formulas. (...) It would in no way disprove or contradict
the hypothesis, or weaken our confidence in it, if it should turn out
that the billiard player had never studied any branch of mathematics
and was utterly incapable of making the necessary calculations...”
(Friedman and Savage [1948, 298], italics in original).
Of course, modeling a billiard player as if he solves a system of equations yields
more accurate predictions when one considers excellent players than for a notorious beginner. Thus, the model would arguably be more fruitful to describe games
at championships than at the average elderly home.
3
Economics and psychology
A similar argument goes through for using the standard assumptions behind
economic man. Modeling an agent as if he can solve an optimization program is
likely to yield better predictions for some agents than for others. One is in particular tempted to think that agents who are actively engaged in market transactions
can reasonably be modeled this way. The market would punish those agents who
do not behave as if they solved the optimization programs properly. These agents
would realize losses, a situation that cannot be sustained for a very long time.
Or at least they would on average make less profits and be competed away. According to similar logic, the market would leave no room for other than purely
self-interested agents. Setting a price that is perceived as fair but is not competitive, induces losses as well, leaving the opportunistic agents in the market.
We would thus be left with the (as if) maximizing and selfish agents as market
participants. Since economics is in particular oriented towards studying markets,
the ”as if” assumption seems innocent in this field. In fact though, neither one of
the above claims is necessarily true. Both non-maximizing and boundedly selfish
players can survive market forces. Arbitrage opportunities are limited, apparently
even in the realm of financial markets (Mullainathan and Thaler [2000]). Learning
by agents may over time lead to the competitive equilibrium, but learning itself
is often a costly and slow process.
All in all, the defense of modeling agents as if they are maximizing selfish
agents seems unwarranted on many occasions. Moreover, to the extent that the
market will surpress non-maximizing or unselfish agents, it is still interesting to
study what kind of heuristics (like simple rules) agents would use otherwise, or
with what kind of sentiments they are equipped. Only then is it possible to judge
market efficiency in comparison to other institutions that bolster these sentiments
more than the market does (Rabin [2002]).
1.3 Economics and psychology
Assuming less intelligent and selfish agents than usual is not always straightforward or even useful. Sometimes the agents make mistakes but is it not well
understood how they reason or why they reason other than is assumed. At other
times, the mistakes are just too small to consider. It also happens that agents
make mistakes but over time converge to the predicted equilibrium. To illustrate:
in an experiment by Nagel [1995], individuals had to state a number in the in-
Chapter 1. Introduction
4
terval [0, 100]. The one closest to two thirds of the average received a price. It is
easy to see that in equilibrium all individuals choose 0. In the experiment, this
was not the case. In the first round, the mean and median were around 33. The
mean number did converge to the equilibrium thereafter.
Many times, however, agents make ’mistakes’ which are predictable, important,
and for which there are good explanations. In those cases, it makes sense to model
agents as psychological man instead of economic man.
In this dissertation the consequences of various psychological sentiments are
scrutinized. In order to place the chapters in a broader framework, it is useful
to consider the categorization by Rabin [1998] and Tirole [2002]. They survey
the literature that departures from the economic paradigm. I briefly discuss this
categorization and some of the interesting contributions in the literature so far.2
1.3.1 The economic paradigm
Tirole [2002, 634] summarizes the economic paradigm as follows. The individual
is thought of as ”maximizing at each instant t over some action set At the expectation of the present discounted flow utility of consumption uτ (cτ ) given the
information It he has accumulated prior to date t”:
max E
At
hX
τ ≥t
δ
τ −t
¯ i
¯
uτ (cτ )¯ Iτ .
(1.1)
Disentangling the maximization program, the following elements can be distinguished: the utility function, beliefs, discounting, and optimization. For each of
these elements, violations of the usual assumptions are identified. In the remainder of the section some of the many contributions are highlighted. Some of them
are elaborated upon more in later chapters.
1.3.2 Preferences
The starting point of most economic analysis is a concave utility function that
is only a function of individual i’s own bundle of consumption goods: ui = u(ci )
with u0 (ci ), u00 (ci ) > 0. There are many indications that this functional form does
not capture many subtleties that enter the agent’s decision. Both the functional
2 The
most extensive survey is that by Rabin [1998]. The survey by Tirole [2002] includes some of the most
recent contributions to the field.
Economics and psychology
5
form of how much utility is derived from consumption as well as the idea that
only own consumption matters do not reflect the true complexity of real behavior.
u00 (ci ) S 0
To begin with, consider individuals’ choices isolated from interaction with other
agents. The existing evidence from experiments suggests that the utility function
contains a reference point at the status quo. For gains, the utility function is
indeed concave as is usually assumed, implying risk aversion. For losses, however,
the utility function turns convex, implying risk seeking behavior (see Tversky
and Kahneman [1992]). Individuals are loss averse in the sense that a small loss
compared to the status quo is not outweighed by an equal gain. These properties
together, and some more, are elements of what Kahneman and Tversky [1979]
dubbed prospect theory, now one of the most well known theories.
ui = u(ci , cj , ·)
It is often suggested that people derive not only pleasure from own consumption,
but also from the happiness of their friends and relatives, from the fact that
other people behave nicely towards them, from possessing more wealth than their
neighbours, etc. In short, they have social preferences: they care about the payoffs
of others (Charness and Rabin [2002]).
Consider for example the following series of experiments. In the dictator game,
one person has the power to distribute a sum of money over another person and
himself. He can do so in any way he wishes. Often, the other person is anonymous.
The observation that the player who divides the money (the proposer) usually
does not keep all the money to himself points to a notion of fairness. Apparently,
agents derive pleasure from being fair and it is not regarded as a fair distribution
to keep all the money.
In an extended version of the game, the ultimatum game, the proposal of the
player is not immediately implemented. The second player (the responder) gets a
chance to either accept or reject the proposal. If he rejects, nobody gets anything.
If he accepts, the sum is divided as proposed. After a rejection there is no second
opportunity to make a proposal (at least not with this combination of players).
It can be expected that the responder will not always accept positive but small
Chapter 1. Introduction
6
enough offers. Since he may dislike unequal distributions, he may decline offers
where he gets too little and the other too much in his view. This is indeed found.
Interestingly, however, we can learn more from this game. As it turns out, it is
not only the (inequality of the) distribution that the players care about, but also
the intentions of players, something conjectured by Rabin [1993] in a theoretical
paper. To see the relevance, consider the following special case of the ultimatum
game taken from Falk et al. [1999]. In this game, the proposer can only choose
between two possible offers. Either he chooses the offer (8,2) (keeping 8 for himself
and leaving 2 for the responder) or the offer (x, y), where x and y are varied among
different treatments. In the first variant (x, y) is set to (8, 2). This means that
the proposer has no other choice than to propose (8, 2). It turns out that this
offer is rejected in about 20 percent of the cases. In a second variant, (x, y) is set
to (5, 5) giving the proposer the opportunity to split the sum of money exactly
in two. From the proposers who offered (8, 2) in this treatment, no less than 45
percent is rejected. This, despite the fact that given the proposal, the distribution
is identical in both treatments. The natural interpretation is that in the latter
case, responders were angry because the proposer did have the option of an equal
split, but he decided to go for the unequal distribution anyway.
Collecting data from experiments, a preference for an equal distribution and
reciprocal behaviour (that is, rewarding nice behaviour and punish stingy players)
now seems a robust finding (Fehr and Schmidt [1999], Bolton and Ockenfels [2000],
Charness and Rabin [2002]). It is nonetheless also an established phenomenon that
agents try to distinguish themselves from others rather than trying to become
equal. Worries about status is one of the most recurring patterns in all cultures
(see Wright [1994], Van Kempen [2003], and chapter 3). Inequality aversion and
status seeking behaviour need not be mutually exclusive if status is rewarded
for making the distribution more equal. Often, however, they are. To my best
knowledge, conditions for when status seeking behaviour prevails over inequality
aversion are unknown, but would be interesting for future research.
Finally, a recurring theme in the recent literature is that people are intrinsically
motivated (Deci and Ryan [1985]). Intrinsic motivation refers to an inner state
of satisfaction from being involved in an activity. The activity itself is rewarding. This contrasts with the idea that people only work for monetary (extrinsic)
rewards or in a controlling environment. This partly explains why people supply
voluntary labour or work harder and longer hours than is expected from them.
Economics and psychology
7
This concept is firmly established in psychology, but has not received much attention from economists (Frey [1997b]).
Stability of preferences
A final note on the utility function concerns the stability of preferences. A sizeable
body of research shows a picture of a remarkably labile nature of preferences (see
Slovic [1991]). Illustrative of this are the following well-known examples.
First, there are endowment effects: once goods are part of one’s endowment, the
valuation immediately increases sharply. This effect is present even if the subjects
are made familiar with the object on beforehand, thereby excluding learning
arguments as an explanation (Loewenstein and Adler [1995], Thaler [1980]). In
the experiment by Loewenstein and Adler [1995] for instance, subjects indicated
their preference for a mug, and based on this the predicted mean selling price was
$3.73. A few minutes later, when the participants were actually given mugs, the
mean selling price increased to $4.89.
Secondly, framing effects take place: the choice of agents is sensitive to the way
that a choice problem is formulated. For example, the valuation of a gamble is
sensitive to whether the outcomes are framed as gains or losses relative to the
status quo (Tversky and Kahneman [1992]).
Third, individuals adjust to the state they are in. Such treadmill effects are
discussed for instance by Kahneman [2000, 686]:
”Anyone who has bathed in a cool pool, or in a warm sea, will recognise the basic phenomenon. As one adapts, the experience of the temperature of the water gradually drifts towards ”neither hot nor cold,”
and the experience of other temperatures changes accordingly. A temperature that would be called warm in one context may feel cool in
another.”
Another instability is caused by the simple fact that individuals have difficulties
in remembering how they felt about something. Retrospective evaluations of past
utilities are known not always to be reliable (Kahneman [1994]). This is one reason
why individuals have imperfect self-knowledge (see Baumeister [1998]). We tend
to forget how difficult it actually was to give up an addiction and therefore try
Chapter 1. Introduction
8
to do so over and over again3 . Right after the struggle of the attempt to quit
smoking, a person may decide never to try again. Yet after a while he tends to
forget and tries to give up his habit again.
1.3.3 Beliefs
I now turn to the second element of the economic paradigm: beliefs. The expected utility functional is the standard framework for decision making under
uncertainty. Agents are assumed to maximize the sum of utilities of outcomes
linearly weighted by the corresponding probabilities that these outcomes occur.
Mathematically, individuals are assumed to maximize:
n
X
pi u(xi ),
(1.2)
i=1
where outcome xi is received with probability pi for i = 1, ..., n, and with a utility
function u(x) over outcomes (see e.g. Varian [1992]).
I already pointed out that the utility function itself does not satisfy the usual
assumptions of global concavity (see Kahneman and Tversky [1979]). But even
if it would, Rabin [2000] shows that within the framework, risk aversion cannot
be sensibly explained. He makes this clear by the following observation. Suppose
an individual turns down a bet that gives him a 50 percent chance of winning
$110, and another 50 percent of losing $100. Then, if he is an expected utility
maximizer, this same individual should also turn down a bet which gives him a
loss of $1,000 with a 50 percent chance, no matter what the possible gain would
be. That even a bet with a 50 percent chance of winning, say, $10 billion and a
50 percent chance of losing $1,000 is turned down by anyone can be said to be
counterintuitive, except perhaps for the credit constrained people.
The Allais paradox is an early contribution showing the limitations of the
expected utility theorem. The paradox lies in the choices of subjects between two
sets of lotteries. Denote by X = (p1 , x1 ; p2 , x2 ; ...; pn , xn ) a lottery X which gives
a prize x1 with probability p1 , a prize x2 with probability p2 , etc. Consider first
the following two lotteries, taken from Kahneman and Tversky [1979]:
3 Likewise,
people are not perfectly able to predict their feelings. “Most people are very suprised to learn
that paraplegics are not always miserable and that lottery winners are not particularly happy” (Kahneman et
al. [1997, 396]).
Economics and psychology
9
A = (0.33, 2500; 0.66, 2400; 0.01, 0),
B = (1, 2400).
Thus, lottery B is degenerate and gives a prize of 2400 with certainty. Out of
these two lotteries, 82 percent prefers lottery B. Consider next the following set
of lotteries:
C = (0.33, 2500; 0.67, 0),
D = (0.34, 2400; 0.66; 0).
Out of these two lotteries, 83 percent prefers lottery C. These choices are, however,
inconsistent with each other. This is, then, the paradox. To see this, note that
preferring B to A implies (normalizing without loss of generality u(0) = 0):
0.33u(2500) + 0.66u(2400) < u(2400),
(1.3)
0.33u(2500) < 0.34u(2400).
(1.4)
or equivalently:
However, preferring lottery C to D implies the opposite of equation (1.4):
0.33u(2500) > 0.34u(2400).
(1.5)
To gain in descriptive power, the expected utility model needs to be refined.
Optimism: π 6= p
Non-expected utility models have refined the expected utility model in the utility
domain but also in the domain of beliefs (see Starmer [1998] for a survey). For
example, rank-dependent utility models assume that probabilities, p, are transformed by a weighting function, π(p). The intuition behind the weighting function
is that individuals do not only pay attention to the specific probability that a particular outcome occurs, but also to the probability of an outcome in comparison
to other outcomes (Diecidue and Wakker [2001]). An individual will for example take into account the probability of getting a certain outcome or something
better.
Calibrating the weighting function it is found that individuals tend to overweigh
both high and low probabilities. This can explain phenomena such as optimism
and pessimism. It also explains why people are at the same time risk-seeking and
risk-averse, in other words, why they participate in lotteries and insure themselves
Chapter 1. Introduction
10
against unforeseen events (see Diecidue and Wakker [2001]). This is predicted by
the theory because the small probability of ending up with the jackpot of the lottery and the small probability that their house burns down are both overweighted.
Overconfidence
Besides being optimistic, individuals are also often overconfident: 90 percent of
adults rate themselves as better than average drivers4 , and 25 percent of high
school seminars rate themselves in the top 1 percent on the ability of getting
along with others (see Baumeister [1998]).
Part of the overconfidence is the result of ignorance of relevant information.
This happens even if the costs of obtaining information is insignificant. Blackwell’s
theorem, on the other hand, says essentially that individuals should never ignore
freely available information. The idea is that actions are based on information.
More information enables agents to design better strategies (Grant et al. [1998]).
The first thing to note is that Blackwell’s theorem holds for expected utility
maximizers. However, I have already argued that expected utility theory does
not give an accurate description of behaviour. If, on the other hand, someone is
not an expected utility maximizer, he should sometimes prefer less information
(Grant et al. [2000], Wakker [1988]).
There are intuitive reasons why individuals may prefer to have less information.
First, individuals may have an intrinsic value of ignorance due to psychological
sentiments such as anxiety, hope, or fear (Grant et al. [1998], Ahlbrecht and Weber
[1997]). Not all people would like the idea to know it when they are terminally
ill. Secondly, individuals may attempt to self-commit. This is very much linked to
dynamic inconsistency (i.e. overweighing the present), something I turn to now. I
come back to the relation between overconfidence and self-commitment in section
1.4.
1.3.4 Discounting
The third element in (1.1) concerns discounting. Suppose you get the choice
between $50 now or $100 in 2 years. Which one do you prefer? A majority of the
adults from a sample report that they prefer to have $50 now. Now consider the
4 Although
this can be in line with the true distribution, the distribution would have to be very skewed.
The second example (ability of getting along with others) is not compatible with any distribution.
Economics and psychology
11
choice between $50 in 4 years or $100 in 6 years. Almost no adult prefers the $50
in 4 years. Are these choices consistent with the assumption that agents discount
the future exponentially?
Exponential discount factors are most frequently used in economics. Thus, as
in equation (1.1), in period t, the flow utility at time τ is discounted by the factor
δ τ −t . This has the following prediction: suppose an individual has the choice
between consumption level c to be received n periods from now, and c0 to be
received n + ∆ periods from now. Then, if he prefers c to c0 = c + x for a given
n, then he should prefer c to c0 for any n. Since c gives present value utility
δ n u(c) and c0 gives present value utility δ n+∆ u(c0 ), he chooses c if and only if
δ n u(c) ≥ δ n+∆ u(c0 ) or, equivalently, u(c) ≥ δ ∆ u(c0 ). Hence, what matters is the
time lag between dates, but not how far in the future they are. If individuals make
their choices in a way that satisfies this property, they are said to be dynamically
consistent.
This is not how people (or some animals) choose (Ainslie [1991]). In the above
experiment the time lag between receiving $50 or $100 is 2 years in both cases.
Thus, if people prefer $100 in 6 years to $50 in 4 years, their behaviour can only
be consistent with exponential discounting if they also prefer $100 in 2 years to
$50 immediately. But the experimental data shows otherwise.
The choices of above are inconsistent with exponential discounting. The choices
of the individuals can be described by a slightly more complicated discount function, called a hyperbolic discount function (see e.g. Laibson [1997]). This is of the
form:
X
δ τ −t uτ (cτ ),
(1.6)
ut (ct ) + β
τ ≥t+1
with β ≤ 1. Hence, the future is discounted by an additional term β relative to
the present. This distorts choices that involve the present and leads to interesting
predictions. For β = 1 the hyperbolic function reduces to the exponential case.
For β < 1 discounting is present-biased.
To illustrate how this functional form describes choices: preferring $50 now to
$100 in 2 years implies:
u(50) > βδ 2 u(100).
(1.7)
Preferring $100 in 6 years to $50 in 4 years implies:
u(50) < δ 2 u(100).
(1.8)
Chapter 1. Introduction
12
In the second inequality, β cancels out because both dates are in the future.
The term β does not cancel out in the first inequality because one of the possibilities is paid out immediately. Obviously, equations (1.7) and (1.8) can hold
simultaneously for a small enough value of β.
Consider now an individual who needs to save so that in five years he can buy
the overly expensive car he wants so desperately. He is predetermined to set part of
his monthly paycheck away on another account. Sadly enough, he is a hyperbolic
discounter. Every month when his paycheck is added on his regular account,
he has to make the trade-off of between saving the required bit or consume it
rightaway. Overweighing the present, he is too tempted to consume. After five
years, the savings account is still empty.
This individual would be much benefitted by having to his disposal a commitment device to save. Fortunately, there are opportunities for him. Illiquid assets
provide a form of commitment (Laibson [1997]). Investments in illiquid assets
that are subject to a penalty for early withdrawal turn the cost-benefit ratio in
favour of saving. Of course, plenty other forms of commitment can be explained
within the same framework, such as self-imposed deadlines, fixing appointments
well in advance to prevent endless postponements, putting the cookies box on the
highest shelf, and moving the (very annoying) alarm clock far away from the bed.
Heroic figures would even tie themselves to the mast in order to self-commit.
1.3.5 Maximization
The last element in program (1.1) is that agents maximize their utility (given
constraints and information etc.). Do agents really maximize? Some studies show
that they do not.
The existence of such non-maximizing agents is nicely illustrated in a study
by Camerer et al. [1997] of the taxi cab drivers of New York City. Rather than
maximizing revenues per hour by working longer hours on a rainy day with many
clients, and shorter hours on less profitable days, they tend to stop working after
reaching a certain target level of earnings. This behaviour is opposite to that
predicted by maximization.
Target earnings and other aspiration levels are found in many other studies
as well (see Diecidue and Van de Ven [2003] for a survey). Managers often do
not choose a project that maximizes expected profits but they rather aim at
reaching a target level (see Payne et al. [1980], [1981]). Similarly, portfolios are
13
Applications
often constructed on the basis of the likelihood that a certain target return will
be met. Some farmers are known to grow safe crops until their subsistence level
is guaranteed, and to grow more risky crops beyond that level (see Lopes [1984]).
In an experimental study by Loomes [1998], it is also shown that individuals
do not maximize payoffs. Individuals had to divide 20 green and 20 white balls
over two bags: A and B. Every bag should contain twenty balls but the individual
could freely decide on the shares of green and white balls in each bag. Now, the
individuals knows that a lottery will take place that selects bag A with probability
0.65 and B with remaining probability 0.35. From this bag, a ball will be randomly
taken and if the ball is green a sum of £20 will be paid to the decision maker.
If a white ball is picked, the decision maker earns nothing. Virtually all models
of decision making predict that the decision maker puts all green balls in bag A,
thereby maximizing the probability of payoff (i.e. 0.65)(see Loomes [1998]). As it
turned out, most people choose to put 13 green balls in bag A and 7 in bag B .
This matches the probability ratio that each bag will be chosen. Apparently, the
individuals used simple heuristic rules (divide the green balls in proportion to the
probability that the bag will be chosen) instead of maximization behaviour, even
though this reduces the probability of gaining £20 with more than ten percent
(from 65 percent to 54.5 percent).
1.4 Applications
Most of the foregoing results were descriptive violations of standard economic
assumptions. However, one would also like to know why we find these violations.
For example, why are individuals overconfident, fair to others, work less with a
higher bonus, or do they ignore information? Exciting insights in these aspects
result from combining some of the above elements.
One of the most powerful results in the recent literature is the assumption
of imperfect self-knowledge. Bénabou and Tirole have used this assumption in
several domains. In this section I highlight some of their ideas.
Carrillo and Mariotti [2000] and Bénabou and Tirole [2002b] explain ignorance
of information by dynamic inconsistency and imperfect self-knowledge. Consider
an agent who has to decide the next period (period 2) whether to undertake a task
or not. Undertaking the task is costly. The benefits of undertaking are received in
period 3 and depend on the ability of the agent. It is assumed that the agent has
Chapter 1. Introduction
14
imperfect self-knowledge and does not know exactly what his ability with respect
to this task is. The individual can, however, learn about his ability in period 1 at
no costs. All the information he collects in period 1 is known in the future, but he
can only acquire the information in period 1 (for example by doing a related test
in this period). Finally, the individual is dynamically inconsistent: his discounting
function is hyperbolic (see (1.6)).
Suppose in period 1 the agent expects a net gain of undertaking the task and
hence that he is willing to undertake it. Normally, he could gain by acquiring
information. It may turn out that he comes to know that he is of low ability in
which case he better refrains from undertaking the task. However, he can also lose
from more information. Consider behaviour in period 2. In this period, the agent
has to decide to undertake or not, at some costs. But with hyperbolic discounting,
he puts extra weight on the current period. He inflates the importance of the
costs and may no longer be willing to undertake the activity. There is therefore
a potential dilemma: in period 1 the agents aims at undertaking, but in period 2
he may reconsider. This creates a ’time-inconsistency region’.
Will he acquire information in period 1? Not if there is a high probability that
the information reveals that he is in the ’time-inconsistency region’ where he
prefers to undertake from the current viewpoint but will reconsider in the next
period. On the other hand, he will not ignore information if the probability is high
that he is of low ability. In that case ignoring information is too costly because
he would undertake the task even though he should not.
Dynamic inconsistency combined with imperfect self-knowledge creates opportunities for ignorance of information for strategic reasons. Ignoring information
keeps self-confidence high, and high-self-confidence makes persistence more likely.
It also explains phenomena like self-handicapping such as drinking before an
exam. Drinking alcoholic beverages has the same function as ignoring information. It makes the outcome of the test unreliable and it therefore reveals less
information. Finally, it learns us something about why individuals tend to memorize achievements in a selective way: successes tend to be recalled whilst failures tend to forgotten. Similarly, successes tend to be attributed to one’s self
whereas failures are likely to be attributed to others (Baumeister [1998]). These
self-serving biases in memory can help keeping self-confidence high, and minimize
the temptation to give up along the way.
15
Applications
In another context, Bénabou and Tirole [2001] combine dynamic inconsistency
and imperfect self-knowledge to explain compulsive behaviour. Suppose you have
incomplete knowledge about your ability to cooperate, in the sense that you
do not know how present-biased your discounting is (you are uncertain about
β in (1.6), a measure of willpower). Sometimes you are involved in short term
relationships, for example in a restaurant where the waiter serves you well in the
expectation of a good tip. At other times you end up in long term relationships. In
relationships that are likely to be short-term of nature, there is a big temptation
to break up the relationship. You are better off leaving the restaurant, that you
will probably never visit again, without leaving a tip. Long term relationships
always pay off if you sustain them long enough. If you have no strong bias to the
present (high β), you will succeed in sustaining the relationship. With a discount
rate strongly biased to the present (low β), you are tempted to give up the long
term relationship making you overall worse off.
In long term relationships you are not exactly sure about your ability to cooperate. If you knew you had a discount rate strongly biased to the present, you
also knew that a potential long term relationship will not last. There would be
no reason to get involved in a long term relationship. If, on the other hand, you
knew that you do not have a strong tendency to overweigh the present, you would
be able to sustain long term relationships with a high payoff.
A forward looking agent may reason as follows. If I manage to cooperate even in
short term relationships, I will recall later that I must have no strong bias to the
present. So I also must be able to sustain a long term relationship. Even though
this person has no direct benefits from cooperating in short term relationships,
he shows this compulsive behaviour to avoid that he will later be afraid to get
involved in long term relationships. This may explain tipping behaviour.
As a last application I discuss crowding-out of intrinsic motivation. As explained earlier (section 1.3.2), many people are intrinsically motivated to undertake an activity. That is, monetary rewards are not always necessary to induce
people to work. It may still be the case that people are not enough intrinsically
motivated. Economists generally solve this problem by implementing an incentive
system based on monetary rewards, to increase motivation. Surprisingly, many
experiments from psychology show that these extrinsic rewards often undermine
motivation. This result extends to the workplace. It is therefore suggested that
monetary rewards crowd-out intrinsic motivation, for a variety of reasons. For
Chapter 1. Introduction
16
example, rewards imply competition between workers which discourages some of
them (see in particular Kohn [1993] for a survey, and Frey [1997b] for an early
treatment in economics).
Intrinsic motivation combined with, again, the imperfect self-knowledge framework can also shed light on crowding out of motivation (Bénabou and Tirole
[2002a]). The elementary idea is as follows. Suppose that in a principal agent relationship, the agent has imperfect knowledge over his own ability to do a certain
task. The only thing the agents gets is a signal, which is correlated with his ability
to do this task, but only imperfectly. The principal wants him to do this task,
and is aware of the ability of the agent. If he knows the agent has low ability,
he reckons that this agent probably has low self-confidence. Thus, in order to
motivate the agent, he has to give a high bonus. However, the agent realizes that
he gets such a high bonus because the principal knows he is of low ability. The
bonus is therefore also a (bad) signal about his ability. Consequently, the high
bonus lowers self-confidence even more.
In the equilibrium of the above game, a high bonus lowers self-confidence. The
bonus motivates the agent to work in the short run. But once removed, the agent
ends up with a lower self-confidence and will be less motivated to work than
before (see Bénabou and Tirole [2002a] and chapters 6 and 7 for details).
1.5 Discussion
”How strange and confusing are people’s conceptions! Sometimes, when
you think about it, you don’t know whether to laugh or cry. Today it
occurred to me that self-sacrificing love is nothing more than an extreme form of egoism.” Alexander Herzen, Who is to blame?
The examples given in section 1.3 show violations of the standard assumptions
in economics. A growing literature combines psychology and economics to formulate alternative assumptions that describe the data better. Here are some of my
(I am afraid unorganized) views on the field of psychology and economics.
First and foremost, combining insights from psychology and economics clearly
leads to interesting results, as section 1.4 and a bunch of other contributions
to the literature make clear. Patterns of behaviour can be explained that could
otherwise not have as easily and realistically been explained by the standard
economic assumptions. Ignorance of information is a good example.
17
Discussion
Second, it is worthwhile to note that in many cases most of economic methodology is being maintained. Bénabou and Tirole [2001] for instance, drop the assumptions of exponential discounting and perfect self-knowledge, but retain the idea
of maximizing agents playing Bayesian equilibrium strategies. Likewise, altruism
can still be modeled using the individualistic approach of economic methodology.
Perhaps literally (as suggested by Herzen) but at least in the ”as if” sense.
However, some disclaimers are in place. Some of the examples to illustrate
violations may be constructed for this purpose or at least constructed to make the
violations most visible. It is not always made clear how sensitive the experimental
results are to variations in the payoff structure. This makes it difficult to generalize
the results.
Moreover, the economic man can still be useful in a (conditionally) normative
way. Even if it does not describe people how they actually behave, it is still a
valuable framework for analyzing how people should behave in order to achieve
certain ends. Thus, I tend to agree with Luce and Raiffa who relatedly discuss
the use of game theory:
”It is crucial that the social scientist recognise that game theory is
not descriptive, but rather conditionally normative. It states neither
how people do behave, nor how they should behave in an absolute
sense, but how they should behave if they wish to achieve certain
ends” (cited by Zwick et al. [1999, 7]).
The same can be said about the economic paradigm, which gives directions for
how to behave conditional on agreement with the underlying assumptions.
Furthermore, it is not always useful to make more realistic assumptions about
individual behaviour. In many cases, though certainly not all, making more realistic assumptions drastically increases the complexity of analysis and this is not
always outweighed by an increased accuracy of predictions. Sometimes, it does
not matter at all. Fehr and Schmidt [1999] have showed that even though agents
behave as inequality averse in some experiments, in other experiments choices
are exactly as predicted by standard economic assumptions (i.e. selfish). Smith
[1962] showed experimentally that in a double auction market, prices converge to
the competitive equilibrium.
Of course, it is still interesting to know when sentiments like inequality aversion
affect outcomes and when they do not, so this argument does not imply that
Chapter 1. Introduction
18
psychological sentiments should immediately be disregarded. What it does mean
is that, depending on the context, it is sometimes no sacrifice not to include
such sentiments in the model. A bit paradoxically, more work on psychological
sentiments is needed in order to know when including these sentiments is not
needed.
Relatedly, one has to bear in mind that, taken as an ”as if” approach, the economic paradigm is still a rough approximation of reality. In this context, Roth
[1995] rightly remarks that ”To the extent that utility maximization is viewed
as a useful approximation of behaviour, it can’t be easily displaced by counterexamples, since approximations always admit counterexamples” (Roth [1995,
78]). Roth continues by arguing that it is nevertheless still valuable to know the
conditions under which the approximations break down. Is the Allais paradox
(discussed in section 1.3.3) an anomaly and sensitive to the parametrization, or
can it be generalized? In this thesis, I have tried to focus on cases where such
breakdowns occur and where it seems to me that a rough approximation does not
suffice.
It is also worthwhile to note the following. It seems that many psychological
phenomena have two sides: an intrinsic value and a strategic role. Donations to
charity are made out of love, but also to gain approval. The balance between
those sides is a delicate matter. Assuming an intrinsic value for a sentiment often suggests that a shortcut approach is taken, and that the more fundamental
motivations are ignored. Or, as Güth [1995, 342] puts it:
”Very often this [explaining anomalies] is done by including additional
arguments of utilities (...) Doubtlessly a lot can be learned from such
attempts to explain experimental phenomena, especially when they
are based on well accepted motivational forces. Very often, this type
of research resembles, however, a neoclassical repairshop in the sense
that one first observes behaviour for a certain environment and then
defines a suitable optimisation or game model which can account for
what has been observed.”
However, it is easy to tip the balance too much in favour of strategic reasons.
Something should not be too easily dismissed as an intrinsic value. Evolutionary
forces can result in intrinsic values such as fairness and other emotions (Frank
Overview of the thesis
19
[1985]). Here, research from other disciplines, notably psychology and biology, is
in particular useful.
In this thesis, and more generally in the field of economics and psychology,
there is a relatively intensive interaction with experiments both from economics
and psychology. The advantage of these experiments are that situations are well
controlled, and that they give much more insight in individual behaviour than
aggregated data. Because of its importance, it is necessary to be aware of the
shortcomings. Therefore, as a final consideration, I would like to point out some
of the limitations of experimental economics.
First, experimental results are sometimes very sensitive to the framing and
wording, and hence one should be cautious in generalizing the results. Secondly,
it is by no means obvious that results can be directly translated into out-oflaboratory situations. Being fair in the ultimatum game is not the same (and
certainly does not imply) that these people are also fair in ”comparable” real life
situations. Thirdly, subjects often have relatively little time to learn the game
and understand the consequences. Experiments quickly become too complicated
to be understood within the available time frame5 . Time constraints pose a natural limit on the complexity of games. Of course, in real life there is also not always enough opportunity to learn, so this argument does not always hold. Fourth,
experiments often use a relatively small sample and are not often replicated (Rubinstein [2001]). The latter is due to the fact that replications are unlikely to
be published. Thus, although experiments have the advance of creating nicely
controlled situations, and provide us with microdata, their shortcomings should
be reminded. In this thesis, I have tried to borrow evidence from experiments
which results seem robust, and otherwise to mention where more replications and
investigations are welcome.
1.6 Overview of the thesis
1.6.1 Main themes
This thesis considers various psychological sentiments that are implemented in
economic models. The purpose of this is to enrich economic models to account for
5 It
is therefore no surprise that the results in experiments are sensitive to things as whether or not a payoff
table is provided in the instructions (Charness et al. [2001]).
Chapter 1. Introduction
20
behaviour that is observed in reality but is normally not predicted by standard
economic models. The two central themes of the thesis will be to explain why
people give and how people react to monetary incentives. The main departures
from (or perhaps better: extensions to) standard economic assumptions are the
inclusion of social preferences, imperfect self-knowledge, and rationalization of
behavior (rather than rational behavior).
Gift-giving
Gift-giving is of interest because at first sight it seems inferior to efficient market
trade but gift-giving is nevertheless still widely observed. In chapter 2, I present
a survey on possible motivations why people give. I argue that two properties
of gift-giving deserve special attention. First, a gift almost never stands on its
own but is almost always followed by a countergift. This is called reciprocity.
Second, gift-giving seems inadequate in the sense that it rarely maximizes the
receiver’s surplus, as a cash gift would, according to standard microeconomic
theory. Chapter 3 then focuses on one possible motivation behind gift-giving
that can explain the two phenomena of chapter 2, namely the demand for social
approval.
Chapter 4 puts the analysis in a more broad perspective by contrasting the
institution of gift-giving to that of the market. It is argued that the market need
not necessarily crowd out gift-giving even though it may be a more efficient institution. Chapter 5 takes an even more positive view on gift-giving by arguing how,
when properly designed, the market mechanism may become more efficient if it is
complemented by gift-giving. The focus in this chapter is on the welfare aspects of
labeling. It is argued that the same motivations as behind gift-giving may account
for the willingness of people to pay price premiums for socially desired goods, e.g
environmentally friendly goods. This partly solves the information problem.
Rewards
The second central theme is how consumers react to monetary incentives, where
attention is paid to effects of rewards that are usually not taken into account.
Economists normally assume that more monetary incentives have a positive and
monotonic effect on people’s behaviour: a reward (bonus, subsidy) for an action
motivates people to take that action. The reason is that the focus is normally on
21
Overview of the thesis
the direct impact of rewards on efforts. The direct effect is due to a preference for
money, or more generally, consumption goods, by agents. Then if, for example,
a reward is conditioned on performance, an agent is more willing to make efforts
to obtain the reward. However, there are also indirect effects of rewards. For
example, rewards interact with other motivations (a desire for approval, say), or
it signals information (such as the perceived ability of the agent).
Including the indirect effects of rewards on behaviour has interesting consequences. For instance, there is evidence that the positive relationship between
monetary rewards and behaviour does not always hold as such. Under some circumstances, rewarding behaviour leads to decreased motivation (see e.g. Deci
and Ryan [1985], Kohn [1993]). By examining indirect effects of rewards, better
understanding of the relation between rewards and motivation is gained.
Chapter 3 argues that if people care about social approval, it may well be that
subsidizing gift-giving may reduce gift-giving rather than enhance it. A more positive result is obtained in chapter 6. Here, it is explained why principals may give
a bonus that is not specified in a contract. Normally, in a relationship with a
finite number of periods and no contract, there will be no bonus in equilibrium.
However, things change if it is assumed that the agent is not perfectly informed
about his ability. In equilibrium, a reward may signal high ability, which increases
self-confidence and motivation. Finally, chapter 7 examines the effects of subsidies when people try to rationalize their behaviour, and it is found that higher
subsidies have a less profound long-run effect than smaller subsidies.
1.6.2 Detailed overview
The chapters are roughly organized according to the two themes giving and rewards, although some of the chapters combine these two themes. Chapters 2 to 5
focus on gift-giving, and chapters 3, 6, and 7 focus on the effects of rewards. Of
course, this distinction is a bit artificial, since in a loose sense rewards are also
gifts.
I now give a somewhat more detailed description of the chapters’ contents as
a reading guide. All chapters can be read independently.
Chapter 2 is a first examination of why people give. Many motivations are
possible: altruism, exchange, fairness, signalling, and social approval. A more
detailed look at the properties of gifts reveal that not all motivations are equally
likely as explanations.
Chapter 1. Introduction
22
The most elementary properties are reciprocity and inadequacy. Reciprocity
means that nice behaviour is rewarded by nice behaviour (positive reciprocity),
just as stingy behaviour is being punished (negative reciprocity). It is indeed
observed that gifts are not one-way transfers as is often thought. Most of the time,
gifts are reciprocated by return gifts. By inadequacy I mean that generally gifts
do not maximize the receiver’s utility. According to standard microeconomics
arguments, cash-gifts are preferred to gifts in kind by receivers, but in reality
cash-gifts are relatively rare.
All motivations are scrutinized keeping these elementary properties in mind.
Altruism is likely to play a role for gift-giving, but not in those instances where
gifts are inadequate. Gifts as exchange can only be sustained for a sufficiently
long time horizon. Fairness explains charity to some extent. Fairness does not
easily explain all the experimental data. Social approval can explain some of this,
and also inadequacy and reciprocity (see chapter 3). Signalling explains gifts
for situations with information asymmetries. If a person is not sure about the
trustworthiness of other players, the other player can signal to him that he is
indeed trustworthy by making a gift. It is also possible that a gift signals to the
giver himself that he is trustworthy and that he will be able to sustain long-term
relationships.
The chapter ends with a discussion where I argue for a hybrid explanation. For
example, people give not so much because they are fair, but because they like to
appear as fair and receive approval for being fair. Furthermore, I argue that in
order to design efficient institutions, it is important to know what the motivations
behind gift-giving are.
Chapter 3 examines social approval as a motivation to give in more detail. First
I show that many individuals care about social approval and status. Then I argue
that these two elements together explain the basic elements of gift-giving: reciprocity and inadequacy. The reasoning is as follows. For a gift, social approval
is awarded. For the receiver, this means a loss in status. This gives incentives
to give back (reciprocity). The first giver may want to prevent the receiver from
giving back to keep his status high, and therefore gives in kind, making it more
expensive for the receiver to return a gift. I also argue that part of charity can
be explained by a demand for social approval. This explains why many individuals donate more when their donations become public knowledge, and also why
23
Overview of the thesis
donations tend to be more densely distributed near the boundaries of published
categories.
There is an interesting relationship with the other theme of this thesis: crowdingout of motivation. Rewarding gifts reduces the sacrifice. This is likely to reduce
approval for a gift. Hence, if people are motivated to give because social approval
is rewarded, monetary rewards may demotivate to give because less approval will
be received.
Then, in chapter 4 a macroeconomic perspective is taken on gift-giving. If one
assumes that gift-giving is a result of the desire to exchange goods, then it seems
plausible that the market mechanism will take over all gift-exchanges in the end.
As the market grows in size, it becomes more efficient and gift-exchange becomes
a poorer alternative.
In this chapter it is however argued that gift-giving is not only an exchange
mechanism but also adds symbolic utility to an exchange. For example, approval
is obtained in a gift-exchange relationship as argued in chapter 3. Symbolic utility
is not generated by the market because the latter is an anonymous institution.
This makes that gift-giving will not be crowded-out by the market mechanism.
It is possible that gift-exchange is sustained even though the market mechanism
is more efficient.
In the foregoing chapter it is argued that utility is derived from giving for
various reasons, among which social preferences. This perspective is also taken in
chapter 5. It is assumed that people are willing to pay a price premium for goods
that are produced with methods that have less social externalities. Examples are
goods produced with environmentally friendly production methods, that avoid
child labour, or where fair wages are paid to employees.
Problematic is that consumers cannot distinguish production methods by examination of end products. Hence, they are not willing to pay a price premium
for goods that claim to have less social externalities of production. Producers are
therefore not willing to invest in more costly production methods.
Two ways are examined to improve upon the imperfect information equilibrium: standards and labels. If the government imposes a standard on production
this means that all consumers are forced to buy goods according to this standard.
Labels are certificates with information that are voluntary. This allows discrimination between consumers. Consumers with a high willingness to pay consume
the labeled goods. Consumers with a low willingness to pay consume the unla-
Chapter 1. Introduction
24
beled goods. It is shown that labels can lead to a higher welfare compared to
standards for an interval of consumer heterogeneity. For sufficiently low or high
heterogeneity, standards lead to a higher welfare.
Chapter 6 studies the effects of rewards on self-confidence. This chapter builds
on the work by Bénabou and Tirole [2002a] who found that a bonus can signal
low ability or a high task difficulty, thereby decreasing self-confidence. Their focus
is on rewards that are specified in a contract. By contrast, in this chapter it is
assumed that the outcome is only perfectly observable to the principal. This
makes a contract impossible. It is then shown that a principal may want to give
an unexpected bonus anyway.
This chapter offers an explanation why principals reward unexpectedly. A crucial assumption is that the principal has more information than the agent about
the outcome of the task. Thus, the theory is more likely to apply in relationships
where the agent is in his learning phase: a child who is learning the piano, or
an employee undertaking a task for the first time. Another possibility is that the
agent performs a small task which is part of a bigger project. If the principal can
judge what the individual contributions from all tasks are, then he can determine
whether a specific agent has been successfull or not. The agent himself may not
be able to make a good judgement about the value of his specific project because
there are too many interactions going on.
An unexpected reward signals a good performance, and raises self-confidence.
On its turn, a higher self-confidence increases motivation in the next period.
In this way, the chapter offers an explanation why discretionary (that is, not
contracted for) rewards are sometimes given. The reward brings good news to
the agent, and motivates him to continue.
In the last chapter, chapter 7, I consider changes in preferences. While changing
behaviour by appropriate incentives has been subject to extensive investigation
within the field of economics, changing preferences has been left largely unexplored. Social psychologists, on the other hand, have paid considerable attention
to the formation of preferences, trying to demonstrate that they are not stable.
Nevertheless, the preference changes are in many cases quite predictable once we
take up the idea that people first make a rational choice, and then seek to rationalize their choice afterwards. Hence, in this chapter both incentives and attitude
changes are taken into account.
25
Overview of the thesis
The idea is that people experience an unpleasant feeling (called cognitive dissonance) created by an inconsistency between behaviour and attitudes (you smoke
even though you believe it causes lung cancer). To reduce this unpleasant feeling,
individuals often try to rationalize their behaviour, for example by focusing on
certain arguments congruent with their behaviour or by disregarding information
that is incongruent.
As an application, I consider the effects a subsidy will have on the consumption
of goods that have social externalities, e.g. environmental friendly goods (“green
products”). The main result is that a low subsidy stimulates a positive attitude
change towards the subsidized good but a high subsidy does nothing to the attitudes of people. This fits the experimental evidence. I therefore conclude that
high subsidies are ’too much of a good thing’: they affect current behaviour but
fail to affect attitudes and therefore future behaviour.
2
The Economics of the Gift
”One dollar and eighty-seven cents. That was all.” So the Christmas story by
O’Henry begins. The main character, Della, had, with great pains, been able to
save one dollar and eighty-seven cents for a Christmas gift. As this was in her
opinion not enough, she went to the shops to sell her possession she was most
proud of: her hair. With the money earned, she could just afford a splendid chain
for her husband’s watch, the only object of value he possessed. When the door
opened, her husband stared at her with a peculiar expression, bedazzled from what
he saw. He had just bought her a Christmas gift as well: a set of combs worshipped
by Della. And he had bought it by selling his watch. Two foolish young people had
sacrificed all their treasures for gifts that had no purpose. Were these two young
people foolish or, as O’Henry himself thought, did they give the wisest gifts of all
gifts given?
2.1 Introduction
Historically, exchange has been — and still is — one of the most fundamental objects
of study by economists. It is, for instance, one of the basic ingredients in general
equilibrium theory and modern theories of economic growth. Without exchange,
0I
am indebted to Jeffrey James, Theo van de Klundert, and Sjak Smulders for very useful comments and
suggestions.
Chapter 2. The Economics of the Gift
28
no specialization is possible. Without specialization, it is hard to imagine how
economies would ever grow rich. It is therefore of great importance to understand
the functioning of a society. The logical starting point for that is to understand
how exchange is organized.
If we were to give a very crude historical account of exchange, one could probably distinguish three phases, seemingly characterized by an ever higher degree of
security and efficiency. In the first stage, exchange relied on gift-giving to organize
societies. If we look at today’s primitive societies, we indeed see a heavy reliance
on gift-giving. Since a gift is thought of as a one-way voluntary transference of
property, it is not particularly efficient nor is a full exchange secured. A little
further in history, one would observe barter trade. Still inefficient, it is secure
in the sense that it is a true exchange, not only a one-way transfer. In the last
phase we find the most advanced institution to organize exchange, one that is
ubiquitous in developed countries: the price system. The price system is particularly efficient in allocating goods by avoiding the need for a double coincidence
of wants, something that is not accomplished with barter trade. It also allows for
a much greater degree of specialization.
Seen from this perspective, gift exchange should not be of much importance
in today’s market oriented economies. The extravagance and importance of giftgiving in primitive societies1 is therefore primarily studied by anthropologists,
and not so much by economists. However, viewing gift-giving as a primitive mode
of exchange does not do enough justice to this complex institution. For instance,
it does not explain why the tribal economies which are oriented towards gift exchange have not been destroyed but sometimes even flourished in the presence of
the —supposedly superior— market economy (Gregory [1989]). The ’efflorescence
of gift exchange’ thesis, by which it is meant that gift exchange has not suffered under the impact of market economies2 , is therefore considered as a valid
description of modern exchange economies.
Fortunately, there have been a number of recent contributions by economists
which acknowledge the more complex role of gift-giving in modern market-oriented
economies, be it somewhat hidden in specific settings. Akerlof [1982] for instance,
argues that the amount of time that an employee works in excess of the minimum
1 Camerer
[1988, p. 180].
term is borrowed from Gregory [1989]. He relates it to the impact of colonization which is broadly
interpreted here as the introduction of a market economy.
2 The
The Gift
29
requirement can be seen as a gift. Rabin [1993] considers gift-giving equilibria as
situations where fairness considerations lead to cooperative behavior. More clearcut examples include birthday, business, and Christmas gifts, voluntary labour,
and donations to funds. The amount of these gifts in terms of income is sizeable:
money spent on gifts alone by households already accounts for 3-4% of income
(Prendergast and Stole [2001]). Charity donations make up another 2% of income
in the US (Andreoni [2001]).
In this chapter, I intend to survey the economics literature on gift-giving. Other
motivations besides the wish to accomplish a trade are discussed. This is done
with respect to two recurring themes in the literature. One of them is the claim by
many anthropologists that although gifts appear to be voluntary, they create in
fact an obligation to the receiver to reciprocate the gift. The other is the finding
by sociologists that it is very often the case that gifts in kind are preferred to
cash gifts, something that may be regarded as a bit disturbing from an economics
standpoint. I examine to what extent each particular motivation to give can or
should account for these themes, and whether it is an efficient institution as
compared to the market mechanism.
The setup is as follows. First, in section 2.2 some characteristics of gift-giving
are discussed. Different approaches based on motivations are discussed in section
2.3. Each approach is examined on its potential of explaining the characteristics as
mentioned in section 2.2. A general discussion and some conclusions are provided
in the final section.
2.2 The Gift
”To say, here I am. To do something. To give. This is what it means
to be a human spirit.” Levinas, Ethics and Infinity.
There are probably as many occasions for gift giving as there are relationships.
In addition, each relationship is characterized by its own particular demands on
how the gift is given. Should the gift be unwrapped at the spot? Is even a quick
look into the envelope inappropriate? It is therefore not evident what these different kinds of gifts have in common: a Christmas gift is evidently unlike a business
present, and neither do an end-of-the-year bonus and a charity donation resemble
each other a lot. There are, in my view, nevertheless some essential characteristics
Chapter 2. The Economics of the Gift
30
of gifts. Two of these in particular form the backbone of the discussion throughout the chapter: reciprocity and adequacy. These elements play an important role
in most of the gift exchanges.
Reciprocity
At first sight it seems quite natural that a gift is voluntary in nature. Still, anthropologists stress that although voluntary on guise, factually gifts have strong
reciprocal properties (Mauss [1925], Codere [1950]). One has not even only a
duty to give, but also to receive and to return. The extravagance of gift-giving in
primitive societies is underlined by the fact that a failure of accomplishing one’s
obligations to reciprocate often eventuates in warfare and the loss of dignity.3 It
is therefore often thought that reciprocal behavior is necessarily connected with
gift exchange. Mauss called reciprocity one of the ”human bedrocks on which society is built” [quoted by Arnsperger [2000, 72]. Or according to Binmore [1998,
24]: ”Love and Duty are not the cement of modern societies ... the mechanism is
reciprocity” (his emphases).
According to Camerer [1988] however, it is ”especially misleading to assume
that modern gift-giving must be reciprocal”. It is indeed reasonable not to assume
that it is a necessary aspect. Consider for example the case of blood giving. The
giving of blood is not directed to specific individuals but to an anonymous agent,
as carefully remarked by Arrow [1972]. Gifts or donations of this kind can by
assumption not elicit reciprocal gifts, albeit this not immediately signifies that
non-reciprocity is also unlikely to occur in personal relationships. But consider
the higher effort of workers above minimum firm standards. This is not always
reciprocated by the firm in the form of higher wages or bonuses (see Akerlof
[1982]). If we take this behavior as a gift of the worker to the firm, then reciprocity
is not connected with fairly personal relationships either. The correct conclusion
would be that gifts are not necessarily reciprocal in nature. If we are to explain
the existence of gift-giving, we also have to explain why certain kinds of gifts are
given with a reciprocal intention and why others are not.
Adequacy
3 This
occured for example among the Kwakiutl. It should be noted however that their use of warfare mostly
refers to warfare directed to an individual and not so much between nations. For a detailed description of the
Kwakiutl, see Codere [1950] and chapter 3.
31
The Gift
Consider the following two quotes. According to Camerer [1988, 198]: ”A deliberate cash gift is a polite way of saying, we care about you less”. And Douglas
and Isherwood put it even more to our imagination by writing: ”...in our society
the line between cash and gift is ... carefully drawn. It is all right to send flowers
to your aunt in the hospital, but never right to send the cash they are worth”
(Douglas and Isherwood [1978, 58]).
One wonders why it is so bad to make a gift in cash. Standard microeconomics
arguments tell us that it can never be worse to get money rather than a specific
good. The reason is simple: with a cash-transfer it is in principle always possible to
buy the same good as the intended in-kind transfer. Moreover, if existent, a more
preferred good may be purchased instead. Whenever an in-kind transfer forces
the recipient to consume more of that particular good than he would have done
with a cash transfer, the recipient prefers a cash gift (see e.g. Mankiw [1998]).
Because gifts in kind weakly lowers the recipient’s utility relative to a cash-gift,
I call them inadequate.
Besides the literary example from the introduction to this chapter, there is
ample empirical evidence of inadequate gift-giving. Calculations by Walfdfogel
[1993] show that for Christmas gifts, recipients valued the gift by 10-30% less
than what the givers had spent on them. An extreme example of inadequacy
is found among a tribe in Canada (the Kwakiutl) where some of the gifts are
worthless to the receiver (see also the introduction to chapter 3).
For sake of completeness, I should add that there are some notable exceptions
to the rule that gifts in kind are inadequate. First, it may be the case that the
gift is more expensive for the recipient than for the giver. For example, a souvenir
brought from abroad cannot be purchased from an amount of cash that equals
its retail price, as the recipient would have to incur transportation costs. Another
example is when the recipient has incomplete or imperfect knowledge about his
own preferences, such as when he is not aware that this particular good existed
so that he could never have bought this good with a cash-gift even though he
derives great pleasure from it (see also Camerer [1988]). Finally, a gift can help
recipients to stick to self-control. Thaler [1999] considers the example of a couple
who cannot afford to spend more on wine than $10 on average. To retain self
control, they may decide never to spend more than $20 on a bottle, even though
they prefer to have a $30 bottle occasionally. A bottle of $30 as a gift may be
greatly appreciated since they are able to enjoy a fine wine without giving up
Chapter 2. The Economics of the Gift
32
self-control. Although interesting in its own right, in the rest of this chapter I
assume that gifts in kind are inadequate.
The challenge, then, is to find theories of gift-giving that are capable of unifying
these dimensions of gift-giving. This is the purpose of the next section.
2.3 Motivations to give
Being familiar with the characteristics of gifts, I next review some approaches in
the literature and determine the potential explanatory power of each of them.
The aim of this section is to assess the different, sometimes competing, models of
gift-giving with regard to the characteristics mentioned. In order to structure the
discussion I classify the different models based on their underlying assumptions
with regard to motivation. To that end, I distinguish between exchange, altruism,
fairness, social approval, and signalling.
2.3.1 Gifts as exchange mechanism
Probably the most obvious approach lending support for gift-giving is to think
of gift-giving as accomplishing an exchange between agents. Above all, gifts are
found most profoundly in primitive societies. And indeed, it is not unreasonable
to assume that at least initially gifts served as a way to separate production
from consumption. In this way consumption could be diversified and production
could be increased through specialization. The market economy can in this way
be interpreted as a more efficient way of exchange, one where gifts are replaced
by the use of money. Indeed, Kranton [1996a] argues that this is the case. In her
model, agents choose between reciprocal (gift) exchange and market exchange.
Since the market is characterized by a thickness externality —more agents on the
market reduces search costs— eventually all gift exchange relationships vanish
whenever the market size exceeds a threshold level.
While intuitively appealing, the model of Kranton [1996a] cannot account for
the coexistence of gift exchange and market exchange.4 If contemporary markets
are so efficient as we think they are, why do people still partly stick to gift ex4 This
is not entirely right. The model is able to predict market size for which gift exchange is sustainable.
But the model cannot explain how evolution got us in this equilibrium except for some shocks that can be
responsible for this. If we start in a gift exchange relationship and some agents find it attractive to enter the
market, then the model predicts that all agents enter the market.
33
Motivations to give
change? Is it not just cheaper to buy all goods and services at the market? Of
course, one reason could be that some products cannot be efficiently produced
on the market. Another line of reasoning is provided in chapter 4. There, I argue
that gift exchange contains a social interaction element that is valued in itself by
the trading agents. Quite often there is a need for mutual sympathy and recognition. These are suppressed entirely in the formal anonymous markets usually
studied (Bowles [1998]). But mutual sympathy is rooted in human nature, as is so
breathlessly described in Kropotkin [1904]. Thus workers develop sentiment for
their co-workers and institution (Akerlof [1982]) and gifts ”symbolize and convey
meaning” (Camerer [1988, 181]).
In the terminology of Khalil [1997], gift exchange provides symbolic utility on
top of substantive utility. A good consumed therefore gives its ordinary substantive utility —in a market exchange as well as in a gift exchange relationship — and
on top of that the agent experiences symbolic utility but only if the trade has
been accomplished in a gift exchange relationship.
This symbolic utility has to be explained in somewhat more detail. Let the
valuation ratio refer to the ratio of utilities that one experiences in a gift exchange
relationship and on the market. It is suggested in chapter 4 that the valuation
ratio is dependent on the market size in two directions. First the valuation ratio
tends to increase as the market gets larger. This is so because mutual sympathy
and recognition are lacking in anonymous market exchange relationships, making
sympathy more valuable.
However, there is also a tendency for the valuation ratio to decline. This idea
builds on the literature on cognitive dissonance in psychology. People have a
resistance to change that is lower if more people are supporting a certain view. If
agents have to decide whether to stay in their personal gift exchange relationship
or to enter the market, then the decision to enter the market gets easier with a
larger market size; in essence if more people are supporting the same view.
It is argued that these two opposing tendencies are likely to result in a valuation ratio that is first declining and then increasing in the market size. Under
appropriate conditions, this model predicts that the market can become efficient
enough to attract part of the population. But as the market gets larger, the valuation ratio becomes larger (lack of sympathy becomes more and more oppressing)
and the people who stayed in the gift exchange will decide not to enter the market
after all. They stay even though the market has become more efficient due to the
Chapter 2. The Economics of the Gift
34
larger size. This can be a stable equilibrium, no agent having the incentive to
switch regimes. As a result part of the population is involved in market exchange
and part of the population in gift exchange.
The model described is interesting in itself since it argues that the focus of economics should not be a one-sided inquiry into the market as a possible exchange
mechanism. In addition, the model can explain a number of things mentioned in
the previous section.
First it is able to explain the seemingly inadequacy of gifts by taking sympathy
into account. For example, it can be that the market provides the same good at
lower costs. If people still consume the good within their gift exchange relationship
then this points to an inadequate gift. The reason is that part of the utility is
neglected; symbolic utility. Substantive utility is higher in the market (more goods
at the same costs) but the market provides no symbolic utility. Hence, on net gift
exchange is preferable. If in reality we only look at substantive utility, then the
gift seems inadequate. If we take into account symbolic utility, there is no matter
of inadequacy. Once we take this properly into account we are able to explain the
sustainability of gift exchange.5
Secondly, gifts have an obligatory element to reciprocate. It is even part of the
motivation to reciprocate gifts. If some agent does not return, the relationship
ends and both enter the market.6 As a consequence, an important class of giftgiving, namely charity, cannot be explained by exchange as a motivation to give
as this usually takes place anonymously and without a countergift. Moreover,
the model is not able to explain why some gifts have no reciprocal character or
how we can trust people in short run relationships, something that is resolved in
section 2.3.5.
2.3.2 Altruism
Another motivation for gift-giving, perhaps a more natural one in the eyes of
people from countries with well developed markets, is to consider the idea that
persons have altruistic feelings towards each other. Within the economic method5 There can still be inefficiency in that everybody could be made better off if all people would enter the
market or all would stay in their gift exchange. This is due to the existence of externalities that are present in
the model.
6 This is partly due to the assumed tit-for-tat strategy of the players. But it seems that this or any such
strategy where the cheater is ultimately punished is reasonable.
Motivations to give
35
ology, this can be modeled as individuals having either preferences for the consumption level or the utility level of other individuals. The structure of a utility
function that represents such preferences is given by Ui = Ui (xi , Uj (xj )), where xi
is the consumption level of person i. If person j has altruistic feelings for person
i as well, there is an infinite regress: Ui = Ui (xi , Uj (xj , Ui (xi , Uj (...)). The regress
easily becomes an unbounded process but Becker [1974] shows an example where
it is not. For example, consider the utility function:
Ui = xαi Ujβ .
(2.1)
Then the reduced form of the utility function follows straightforwardly by substitution and is given by:
α
2
αβ
2
Ui = xi1−β xj1−β .
(2.2)
Clearly, this is finite for α ≥ 0 and 0 ≤ β < 1.
Altruistic feelings will take care of a redistribution such that an optimal balance results between personal consumption and consumption of the other. If the
endowment of a particular individual is in his view relatively too high, he can
gain by giving some of it to the others. Existence of gift giving can therefore be
rationalized.
There are several aspects of this model with respect to efficiency that are noteworthy. First, the equilibrium allocation is generally not efficient because neither
one of the players acts like a social planner despite their altruistic feelings. To
see this, consider player i being altruistic towards j but not vice versa. Note that
player i maximizes Ui by setting xi such that:
dUi
∂Ui ∂Ui ∂Uj dxj
=
+
= 0,
dxi
∂xi ∂Uj ∂xj dxi
whereas a social planner would set xi such that:
·
¸
∂Ui ∂Uj dxj
∂Ui
dUi dUj
+
=
+
1+
= 0.
dxi
dxi
∂xi
∂xj dxi
∂Uj
(2.3)
(2.4)
In other words, the social planner counts player j twice: once because player j
has his own utility, and one more time because player i derives utility from him.
In general, this means that the optimal choice of xi by player i differs from that
by the social planner.
Within a family context, Becker [1974] has shown that an altruistic head of the
family internalizes externalities within the other selfish family members by the
Chapter 2. The Economics of the Gift
36
appropriate transfers (the ’rotten kid theorem’). However, this efficiency result is
somewhat special (Bernheim and Stark [1988]). In particular, altruism can create
inefficiencies, such as is the case in the ’Samaritan’s dilemma’. This dilemma
concerns the problem that if a recipient knows that he will be helped out by an
altruist, he has less incentives to, say, self-discipline himself by saving money for
the future (Bernheim and Stark [1988] provide a more detailed discussion).
Another, related, efficiency result is obtained by Kranich [1994]. Suppose some
players have preferences over the entire allocation of the economy, rather than just
one’s own consumption level. This can be due to altruism, but also to a preference
for fairness. In this case, Kranich proves that the set of Pareto-efficient equilibria is
a subset of the set of all gift equilibria. In other words, the equilibrium that results
when agents can freely redistribute endowments need not be Pareto-efficient. This
can possibly be caused by the public good character that gifts can take. One can
think of a case with three players. Would players 1 and 2 both give to a third
player everybody may be better off, but if either one of the players gives then it
is not profitable anymore for the other to give7 .
Let us now consider the question whether altruism can account for reciprocity.
With only one good, the answer is negative.8 If we consider more goods, however,
one can easily see that a recipient may indeed have an incentive to reciprocate.
If the endowments are sufficiently different between persons, they may all gain
by redistributing, very much like the logic of international trade models. With
altruistic feelings, this redistribution may be accomplished without further motivations, since each player gains indirectly by giving part of his or her endowment
to another person who would be made better off.
Concerning adequacy, however, we see that altruism as a motivation to give
is incongruent with giving in kind. Any cash gift would make the recipient better off9 , without changing one’s own consumption, and it therefore necessarily
increases one’s own utility. Hence we conclude that altruism alone is not a good
7 Goldman
[1978] puts less restrictions on preferences and finds that in that case the reverse also holds:
Pareto-efficient equilibria need not be gift-equilibria. It is for example possible that a gift from person 1 to
person 2 may decrease the welfare of person 3 (because he cares, say, about the consumption level of person 1),
hence moving away from a Pareto-efficient situation.
8 This is true because in equilibrium it must be the case that if player 1 gives to player 2, player 2 has
no incentive to give back. For suppose he has, then his utility from giving some of the good to player 1 must
increase. But then player 1’s utility should also increase, since the utility of player 2 increases and his own
consumption as well. This would contradict player 1 playing an equilibrium strategy.
9 With the exceptions mentioned earlier.
37
Motivations to give
candidate to explain most of the situations where gift-giving takes place, as it
cannot explain the widely observed inadequacy of gifts. Still, there is at least one
important situation where altruism cannot be excluded, namely that of charity.
Empirical studies are indeed supportive of the view that altruistic motives lie behind charity, although these studies at the same time demonstrate that altruism
alone cannot be the unique motivation (see section 2.3.4 for more on this).
2.3.3 Fairness
As a third motivation for gift-giving, the focus in this section is on fairness considerations. This approach recently got most attention in the literature. It successfully accounts for a broad range of experimental games by assuming that people
not only care about their own monetary payoffs, but also about the distribution
of payoffs and the intentions that other players have.
To see the power of this approach, consider the ultimatum game. In this game,
the proposer gets to offer a share of a certain amount to the responder. If the
responder accepts, the responder gets the share and the remainder is for the
proposer. May he reject, both players end up with nothing.
The game-theoretic prediction of this game is easy to see if only purely selfish players are assumed to participate. Since the responder is always better off
accepting any positive amount than rejecting it, the proposer offers the smallest
possible positive amount. This prediction is, however, clearly refuted by the data.
Most offers by the proposer are nowhere near zero. Typically, they are between
40% and 50% of the amount of money (Fehr and Schmidt [1999]).
Several authors have suggested that the data can be fruitfully explained by assuming that people are not purely selfish but have more ’social preferences’ (Charness and Rabin [2002], Bolton and Ockenfels [2000], Fehr and Schmidt [1999]).
There are at least two important components to these preferences. Let us first
focus on the first component. The first component consists of a preference for
equality. This explains why proposers give more than the smallest possible offer
in the ultimatum game. They may consider the spread between payoffs too large
if they offer the smallest possible amount. By giving some of it to the responder,
they reduce the inequality a bit.
The same model can also explain reciprocity. In the so called gift-exchange
game, the proposer can give an amount of money from his endowment to the
responder in the first round. The amount offered is then, say, doubled. After
Chapter 2. The Economics of the Gift
38
that, in the second period, the responder has a chance to make a countergift.
In this game, both players can be made better off by a gift and a countergift.
The structure with only selfish preferences leads to an equilibrium with no gifts.
With inequality aversion, Fehr and Schmidt [1999] show that gift-giving can be
explained, something also found in the data.
The strongest evidence that people care about the distribution of outcomes
is in my opinion the result of another variant of the ultimatum game. In the
dictator game, the responder gets again to propose a share to the responder, but
this time the responder has no say in the outcome: every proposal is directly
implemented. It is found that some of the amount of money is still directed
towards the responder (see Bolton and Katok [1998]). This result is particularly
strong as in this game no strategic effects on the part of the proposer should be
present: the responder has no power anyway.
The second component of social preferences is the part of intentions. People
not only care about outcomes, but also about how these outcomes are realized.
This is clear from for instance the game in figure 2.1 taken from Falk et al. [1999].
P
propose (8,2)
propose (x,y)
R
R
accept
reject accept
reject
(8,2)
(0,0) (x,y)
(0,0)
FIGURE 2.1. A variant on the ultimatum game
39
Motivations to give
The game is a variation on the ultimatum game and goes as follows. In the first
stage the proposer can propose the offers (8, 2) or (x, y), e.g. (8, 2) means 8 for
the proposer, 2 for the responder. The values of (x, y) differ among experiments.
In the second stage (after observing the offers of the proposer) the responder can
accept, in which case each player gets the proposed offer, or rejects, in which case
none of the players gets anything.
In the first variant of the experiment (x, y) was set to (8, 2). Note that the
proposer can in this case only make the offer (8, 2). Clearly, if the responder
only cares about monetary payoffs he will accept the offer giving him a payoff
of 2 rather than rejecting and getting nothing. But in fact, in 20 percent of the
cases the offer is rejected. A possible explanation can in fact indeed be inequity
aversion. Accepting would yield a higher payoff but also an increased inequality
between the two players.
This is, however, not the complete story. Another variant was played in which
(x, y) was set to (5, 5). The proposer can in this case offer (8, 2) or (5, 5). The
situation for responders remained unchanged for those whose proposer sticked
to the offer of (8, 2). He still ends up with either 2 when accepting or 0 when
rejecting, and the inequality is the same for him as in the first variant. The
prediction is therefore that if he accepted (rejected) in the first variant then he
should accordingly also accept (reject) the offer (8, 2) in the second variant. But
apparently something did change for the responder because the rejection rate
increased from 20 to 45 percent when (8, 2) was proposed. The explanation of
Falk et al. [1999] is that intentions matter as well. In the first variant proposers
had no choice but to offer (8, 2). What matters is that in the second variant
they could have chosen to propose (5, 5) but they did not. As a consequence, the
bad intentions of the proposer were punished by the responders. Therefore, they
conclude, both unfair outcomes and unfair intentions matter.
Reciprocity is replicated in many experiments, either in the form of rewarding good behavior (positive reciprocity) or by punishing bad behavior (negative
reciprocity). Whether this is due to responders who try to change outcomes or
to responders who try to reward or punish intentions, in both cases the possibilities for cooperation are increased. Recall however also the negative result from
section 2.3.2 that if people have preferences over the distribution of outcomes, a
Pareto-efficient situation need not be reached.
Chapter 2. The Economics of the Gift
40
Less obvious is how inadequacy can be explained by fairness models. If gifts
are a means to reduce income inequalities, then the most efficient way to do so is
to give adequate gifts (unless, of course, players only have the possibility to give
inadequate gifts or nothing). Also, if the intention is to reward good behavior,
it does not make sense to give inadequately since this makes the reward more
costly.
The in my view most problematic part of social preferences as an explanation
for giving behavior is due to a variant of the ultimatum game introduced by Güth
and Van Damme [1998]. In their variant, a third player is added. In the first stage
the proposer proposes an allocation for all three players. In the second stage, the
responder accepts or rejects. The third player remains inactive. One of the main
results is that only marginal offers were proposed to the dummy player. Although
Bolton and Ockenfels [1998] have shown that this is not necessarily incompatible
with their inequality aversion model, it is my personal belief that this experiment
shows that the proposer was not intrinsically motivated by fairness considerations.
I come back to this point in the general discussion at the end.
2.3.4 Warm glow & social approval
Searching for motivations for charity-giving, Andreoni [1989] assumes that people
have a taste for giving. He reasons that, if people have purely altruistic motives
for donating to a public good, they should only care about the total supply of it.
His ”egoists” and ”impure altruists”, on the other hand, do not only care about
the supply of the public good that they donate for, but also experience a ”warm
glow” from having ”done their bit.” (Andreoni [1989, 1448]). Thus in his model
contributions to a public good are made not only for the benefits of public good
supply but also for experiencing the (egoistic) warm glow feeling.
The distinction between pure and impure altruists is a useful one and an attempt is made to test the hypothesis whether or not people donate to public
goods out of purely altruistic motives. Interestingly, it is found that generally
people are impure altruists: part of their motivation is attributed to the warm
glow feeling (Andreoni [1990], [1993], Bolton and Katok [1998]).
Here I go one step further in trying to explain why people get this warm feeling
from giving. The basic hypothesis, for which there is ample evidence, is that
people are searching for social approval. The following passage that is taken from
the Fable of the Bees illustrates this neatly:
Motivations to give
41
”That a man with small skill in physick and hardly any learning,
should by vile arts get into practice, and lay up great wealth is no
mighty wonder, but that he should so deeply work himself into the
good opinion of the world as to gain the general esteem of a nation,
and establish a reputation beyond all his contemporaries, with no
other qualities but a perfect knowledge of mankind, and a capacity of
making the most out of it, is something extraordinary.” (B. Mandeville
[1714, 262].
This passage concerns the, at that time illustery, Dr. Radcliffe, who gained the
general esteem of a nation by donating his wealth to Oxford University. As he
was aware of, making gifts is a way to be approved by others. The warm glow
is therefore due to the social approval received and not so much for the act of
giving itself. That is, the gift is a means to get a warm glow and has no intrinsic
value in its own in this respect.
Since social approval as a motivation to give is the subject of chapter 3, only
the main ideas are briefly discussed here. The important premises are that people
get approval for a gift and they care about this, and that there is a status element
in the approval domain in the sense that people want to get more approval than
others. It is thus argued that people not only want to be admired, but also want
to be more admired than others (see Holländer [1990]).
With these building blocks, chapter 3 tries to account for reciprocity and adequacy. This is done in a sequential gift-giving game with two players. In the first
period, player 1 has the opportunity to transfer some of his endowment to player
2. Player 2 gets a chance to make a transfer to player 1 in the second period.
Both players have a utility function of the form (for current purposes slightly
simplified):
ui = ux (xi ) + us(si − sj ),
(2.5)
where xi is the consumption level of player i, si his social approval obtained, and sj
the social approval obtained by player j. Total utility consists of a consumption
and an approval part. Social approval is increasing in the size of a gift, i.e. a
transfer of x.
Solving this model, one can see that player 1 has an incentive to make a gift to
player 2 in order to gain approval. This has two effects on the utility of player 2.
First, it lowers his marginal utility of consumption. Second, because player 2 is
Chapter 2. The Economics of the Gift
42
now behind in the race for approval, his marginal utility of approval has increased.
Both effects give incentives to player 2 for making a countergift, explaining reciprocity.
In some cases, the countergift of player 2 is appreciated by player 1 because his
consumption increases. However, he also gives in in terms of net approval. If this
latter effect dominates, player 1 may have an incentive to give inadequate gifts. To
see this, note that an inadequate gift has less impact on the utility of consumption
for player 2, resulting in a higher marginal utility of consumption. This makes a
countergift more expensive, resulting evidently in a smaller countergift.
The analysis of chapter 3 also sheds light on the efficiency, or rather inefficiency,
of gift-giving. Because players are involved in a race to be more approved than
the other, gift-giving is generally inefficient. This results in a standard example
of a prisoners’ dilemma-like game. Both would be better off keeping their endowments, but each player has an incentive to deviate from this situation. When
both deviate, they all lose since net approval sums up to zero and resources are
wasted.
Another inefficiency results from the strategic power of player 1. Although he
does take into account the behavior of player 2, he does not take it into account in
a socially efficient way but only insofar as it concerns himself. That is, he ignores
the effect of his behavior on player 2’s utility.
Note also the positive result. In a situation where no explicit contracts can
be written down, and implicit contracts are infeasible, a mutually beneficial exchange can still take place. Due to the need for approval, players give part of
their endowment away. Their gift is related to, but not necessarily dependent on
whether other players will give something back. The desire for approval does not,
of course, guarantee an exchange whenever there are mutual benefits possible,
but at least creates some opportunities.
It is not immediately obvious how the social approval approach is capable of explaining charity since the benevolences of these gifts are most often unobservable
to the recipient. Recall however that social approval was obtained from the fact
that a gift was made, not necessarily to the person who shows approval. Surely
it is often socially approved to donate to developing countries by the people from
the developed countries. Similarly, you can give blood which is anonymous for
the receiver, but by telling your friends of your act it is not a truly anonymous
gift to everybody. Charity funds seem to acknowledge this aspect by helping to
43
Motivations to give
make donees visible. This can, for instance, be done through the provision of coffee mugs or label pins with the funds’ name, or by publishing the names of the
donees. There is indeed evidence that some people are sensitive to such actions
(see Harbaugh [1998], Andreoni and Petrie [2000]) and in addition that this is
related to the need for social approval (Satow [1975]).
2.3.5 A gift as signalling device
As argued in section 2.3.4, there is more to giving than altruism alone. In chapter
3 I elaborate in much more detail on social approval as a motivation to give, trying
to make a case for it. There certainly seems something to it, as it explains many
of the characteristics of gift-giving for a wide range of circumstances, including
”anonymous” donations. However, as it stands, the model misses one important
element: why should one give to get approval from others?
Social approval may be an important motivation but is perhaps only a derivative of even deeper more fundamental desires. It seems easy to come up with
candidates for which one could get approval: status, being kind, honest, trustworthy, etc. However, it remains to be seen why a gift is necessary to get approval.
Why does the recipient not approve of me without a need of giving? Of course,
one can argue that the gift itself is approved. A gift itself is a kind act indeed,
approval worthwhile. Another interesting line, taken up in this section, would be
to assume that there is an information asymmetry which can be partly or completely resolved by making a gift. The recipient does not approve of me because
he cannot tell what kind of person I am without making a gift. For example,
being fair may be approved by others. But it is not obvious to judge who is fair
or not from the outside. Still, a person who gives may signal that she is fair just
by doing this. Because information asymmetries are so common in everyday life,
Barkow [1989, 100] has even coined humans ’impression managers’ rather than
decision makers.
There are many more possibilities for gifts having the role of signalling something, including income, fairness, and even your own personality. This section
reviews the role and scope of gifts as a signal. I start with a relatively detailed
description of an early contribution by Camerer [1988]. This makes clear how a
gift can be used as a signal. The other variants are then more easily understood.
It should be stressed beforehand that gifts as signals need not be contrary to the
Chapter 2. The Economics of the Gift
44
other motivations mentioned, but may rather be complementary. I come back to
this, together with a general discussion, in the next section.
A gift signals trustworthiness
To enhance cooperation, it is necessary that agents trust each other. In section
2.3.1 it is already being discussed how a gift exchange can take place. There, the
agents could trust each other to complete an exchange by the threat of losing
the trading partner in future exchanges. This is only possible insofar as agents
are not too impatient and if the time horizon is sufficiently long. Camerer [1988]
has shown how a gift itself can signal trust, making even short-run cooperation
feasible.
The essential mechanism is the following. It is assumed that there are two
(groups of) players. Exchange can be realized between those groups. Each group
consists of two types of players: trustworthy players and cheaters.10 The fractions
of these types in the groups are known but the individual type cannot be observed
directly. The main difference between the types is their payoff. Trustworthy players are resistant to cheating. They would feel ashamed if they did, lowering their
payoff. Cheaters on the other hand find it profitable to cheat, they have no feelings of shame or guilt whatsoever. The problem now is that an agent cannot know
beforehand if his trade partner can be trusted. If he is of the trustworthy type
the deal will work out fine, but if he cheats, the payoff will be considerably lower
for the befooled agent than if he would not have traded at all.
What does the model predict? One result is that if the fractions of honest
players in both groups is large enough, then the honest players will trade at
the risk of being cheated. A more interesting result however is that even if the
fractions of honest players are low, trade can still occur. The chances of meeting
an honest trade partner are low, but if there is a possibility of giving a signal of
trustworthiness then this does not need to be so much of a problem. The signal
is to make a gift. This strategy can be explained as follows. If the fraction of
cheaters players is high, then without gifts nobody would be trading. The payoffs
are in this case not very large, but trading would on average be even worse for
the honest players. Now, if a honest player makes a gift of a size that a cheater
10 Nothing
depends on the assumption of two different groups. One may also interpret them as two single
players who are with some probability a cheater or a trustworthy person.
Motivations to give
45
is not willing to make because it is not profitable enough for the latter, then this
is a signal to the other player that this person is honest. And if both players
are honest, then they can trade. In principle the cheater can make a gift as well,
but if the gifts made by the honest person are expensive enough, the cost of
the gift does not compensate for the payoff by cheating.11 If such a separating
equilibrium exists, honest players can signal the trustworthiness of trade partners
by inspection of the size of the gift received.
Example. As an illustration of the above, consider the following example. There
is a honest player (H1) who wants to trade with another person on the market.
There are two players that he can trade with but he doesn’t know which one of
them is the honest (H2) and which one is a cheater (C) (he meets each player
with probability 0.5). He has to decide whether to invest (I) or not (N ). After
investing or not, the trade partner makes a decision. Players that do not meet a
trading agent have a payoff 0. The rest of the payoffs are as in the matrices below.
H2
H1
I
N
I
6/6
5/ − 10
C
N
−10/5
0/0
H1
I
N
I
N
6/1
−10/2
5/ − 10
0/0
As one can see, both the honest and the dishonest players are worst off when
they invested but their trade partner did not. Moreover, both types prefer not
to invest if the other doesn’t. The difference is that the honest player prefers to
invest if the other invests whereas the cheater prefers not to invest if the other
invests. (An economic example may be two persons trading where cheating is
beneficial for both in pecuniary terms, but where honest players have a sense of
guilt outweighing the pecuniary payoffs and the cheaters have no sense of guilt.)
It is readily seen that the dominant strategy for the cheater is not to invest.12
What should H1 do? If he invests, he meets with probability 1/2 H2 who then
also invests and with probability 1/2 C who does not invest. His expected payoff
1 1 Thus a one-shot Prisoner’s Dilemma (PD) game does not satisfy the assumptions needed for a separating
equilibrium. A necessary assumption is that honest players gain by cooperating with another honest player and
lose by cooperating with a cheater. By the structure of a PD-game, cooperating is never a best-response no
matter what the type of the other player.
1 2 Mixed strategies are not considered here as they are not equilibrium strategies (see the appendix in Camerer
[1988]).
Chapter 2. The Economics of the Gift
46
is therefore −2. His expected payoff by not investing equals 0. As a result, the
honest player will not invest and consequently H2 experiences expected utility of
0 by also not investing. What H2 can do is to make a gift to player H1 before
H1 decides.
Suppose that he decides to give an amount of 3 to H1. The net payoffs are then
given in the left part of the matrix below:
H2
H1
I
N
I
N
9/3
−7/2
8/ − 13 3/ − 3
C
H1
I
N
I
N
9/ − 2 −7/ − 1
8/ − 13 3/ − 3
If C does not do the same then it is obvious for H1 what to do. Now he knows
that if he gets a gift after meeting his trading partner, then the other is honest
and so he should invest. Note that the expected utility of H2 is now equal to 1.5
which is still an improvement for him even considering the costs of the gift. What
remains to be shown is that the cheater does indeed not make a gift. Consider the
right part of the matrix above. If the cheater makes the gift of 3, then whatever
the strategy of H1 is, he is worse off than the payoff of 0 when he didn’t make
the gift. The best thing he can do is therefore indeed not to make a gift. ¤
Obviously this model is able to explain the existence of gift giving. It can also
account for some other aspects of gifts mentioned in section 2. It can explain
one-sided gift giving. If the fraction of honest players in group 1 is large and
if group 2 consists mainly of cheaters, then an equilibrium can be that honest
players of group 2 must make a gift to signal their honesty, but for group 1 it is
not necessary to give. The extreme case is where all players in group 1 and only
one player in group 2 are honest. Obviously the players in group 1 do not have to
give to signal whereas the player in group 2 does have to give. This can explain
why gift-giving is not always reciprocal in nature and also when it is: if in both
groups the number of cheaters is relatively large.
Additionally, with a slight modification the model is able to explain inadequate
gift-giving. Recall that the gift must be large enough to make it unprofitable for
the cheaters to give. But if gifts are adequate, under some circumstances it can
still be profitable for cheaters to give since they benefit a lot from the gift they
receive. This would make it impossible to distinguish the honest players from the
Motivations to give
47
cheaters. Whenever the cheaters find it relatively more profitable also to signal,
gifts should be more inadequate or else they fail in their aim.
Example (ctd.) Suppose for simplicity that the honest player H1 in the above
game also makes a gift. This does not change the strategy of the cheater since
both his payoffs from playing N or I are now increased. The adequacy also does
not matter in this case since it does not change the strategy. But suppose that
another pre-stage is constructed. In this pre-stage, some entering costs must be
paid. If the entering costs, T , are paid then the rest of the game is as in the above
example. If the entering costs are not paid by a player, then he is not allowed to
play the second stage and both will not invest. The purpose of the entering costs
in the pre-stage is that the adequacy of the gift can now influence the strategy
that will be played by the cheater. To see this, consider a gift size of x. Only a
fraction δ of the gift adds to the payoff of the other player. The parameter δ is
a measure of the adequacy of the gift. The cheater now has to choose to pay the
entry costs or not, and if he pays the entry costs then he has to decide whether
to give or not. We know that in equilibrium, once entry costs are paid, he will
not give (otherwise the signal is useless). But he may still pay the entry costs
and then collect the possible gift of the other. This would give him an expected
payoff of .5(−T + δx). To prevent him from doing this, the entry costs must be
such that it is not profitable for the cheater to enter the second stage in the first
place: T > δx or δ < T /x. As a result, the adequacy may not be too high. ¤
There are some other models in the same spirit where gifts are taken to be signals
of the willingness to cooperate. Carmichael and MacLeod [1997] derive Nashequilibria where inadequate gift-giving signals the right intentions for long-term
cooperation.13 Bolle [2000] presents a similar model as Camerer and explicitly
derives how adequate gifts should be. Kranton [1996b] derives a strategy for
the formation of relationships by incurring a cost at the beginning of a new
relationship and gradually increasing the level of exchange. In Iannaccone [1992]
it is tried to explain sacrificial behavior within social clubs. Again, sacrifices are
inadequate gifts that signal the good intentions of the players. By demanding a
gift from members of a social club that offers a good which is anti-congestible14
free riding behavior is prevented.
1 3 In
their setup gifts are necessarily inadequate.
indicates that each member’s participation increases benefits for the other members,
contrary to congestible club goods where the benefits decrease with larger utilization of others.
1 4 Anti-congestible
Chapter 2. The Economics of the Gift
48
A gift signals income
People care about consumption goods and a gift may help them to cooperate
and exchange goods. A gift need not necessarily signal trustworthiness though.
To see this, note that people not only care about consumption but equally about
status (see section 3.2.1). One main source of status is, no doubt, the wealth of
a person. The exact amount of wealth that a person possesses is however often
not directly assessable. Glazer and Konrad [1996] show that the level of wealth
can be demonstrated by making gifts. The mechanism is like above. Both the
poor and the rich care about status. For the poor, however, the marginal utility
of status is lower relative to that of consumption. A gift is then a credible signal
of a certain amount of wealth since the poor are not willing to spend as much
resources on status enhancing activities.
A gift signals what you know about the recipient
Also of interest is the idea put forward by Prendergast and Stole [2001] that
a person derives utility from knowing how well they are believed to know the
preferences of the other. This can be an important element in friendships. Besides
this, they assume that people are altruistic. The altruistic motive is a reason to
make a gift to the other. The interesting aspect is what kind of gift they give. The
choice is between a cash gift or a gift in kind. By making a gift in kind he reveals
to what extent he knows the recipient’s preferences. If the good is desired by the
recipient, he will believe that the giver was aware of his preferences, something
that the giver appreciates. A utility loss arises when the gift was improperly
chosen, making the recipient believe that the giver is unaware of his preferences15 .
Suppose that the giver is not completely informed about the recipient’s preferences. If he now gives in kind, he risks a loss of utility because he may end
up buying the wrong gift, deriving no utility from knowing the recipient’s preferences and because the recipient’s utility is lower than a cash gift. If he believes
he knows the recipient’s preference quite well, he may nevertheless be willing to
take this risk. Hence, a gift in kind signals the belief of the giver that he knows
15 A
somewhat related model is that by Ruffle [1999]. Instead of deriving utility from knowing the recipient’s
preferences, he assumes that utility is derived from surprising the recipient, e.g. by making a larger gift that was
expected. This makes it a so called psychological game because beliefs enter the payoff function. Interestingly,
it explains why you should not observe wish lists all the time since it would exclude the possibility of surprise.
Motivations to give
49
the recipient’s preferences. Notice that the story at the outset of this chapter fits
this interpretation particularly well.
There are also situations in which the information asymmetry concerns the
payoffs for the giver. In a principal-agent relationship, it often happens that the
principal is more informed than the agent about the difficulty of a task or the
ability of the agent (Bénabou and Tirole [2002a]). A variant of this is studied in
chapter 6. There, the principal observes the outcome of a task, in essence whether
it has been successful or has resulted in a failure, whereas the agent only gets to
know this information through an imperfect signal. Neither the principal nor the
agent have any direct assessment possibilities of the agent’s ability. It is shown
how a gift (a bonus in this case) by the principal may signal a success, which
indicates that the agent is more likely to be of high ability. A gift then stimulates
the agent to continue making efforts, because he is more self-confident after a
reward.
A gift reveals your self
So far, the informational asymmetry existed in the receiving party having imperfect information about the giver. Relatedly, the giver may have imperfect knowledge over his own preferences. Although the human organism has the capacity to
be conscious about its self, it is by no means the case that each person is totally
aware of his personality. It is also the case that the self is not directly perceived
but built up through experience about one’s own behavior or reflection by others.
For example, Baumeister [1998, 684] writes: ”Consider what is involved in knowing that you are shy. You notice that you always get nervous in the presence of
others, and you prefer to avoid large social gatherings and meeting new people;
these observations permit the conclusion that your are shy”. You can only know
that you are shy through encounters with others. In this way, you learn something
about your self by reflecting on your own behavior.
With imperfect information about the self, there is room for ’self-reputation’.
Bénabou and Tirole [2001] provide an interesting application. Their assumption
involves imperfect knowledge about one’s own willpower. A person may have a
strong or weak willpower, but his type is ex ante unknown to him. In a repeated
setting, past behavior may (partly) reveal the type. By not giving in to temptations today, the individual may later draw inferences that he must have strong
willpower.
Chapter 2. The Economics of the Gift
50
Within this framework, Bénabou and Tirole try to explain tipping behavior.
The individual may get involved in profitable long term relationships. Cooperation in many of these situations would be beneficial. However, the individual
also has short term interests conflicting with long term cooperation. Only if the
individual can think of himself as having the willpower to sustain long term relationships will he get involved in these. The individual may therefore wish to
show strong willpower in short term relationships as well, to signal to himself
that he can sustain long term relationships. Tipping can, according to Bénabou
and Tirole, be regarded as a compulsion: the individual is so motivated to signal
a strong willpower to himself that he is even willing to show cooperativeness in
situations where any direct long term gains are absent.
A gift reveals a strategy
Whereas in the above references signals serve to reveal one’s type, there is also a
variant of signalling that reveals one’s strategy. Here, it suffices to consider only
one type. Consider the battle-of-the-sexes. The essence of the game is that both
players benefit only by choosing the same strategy. Each player has a preference
for one strategy over the other and these preferences are different for both players.
However, choosing different strategies makes both players worse off. Now suppose
that both players can actually make a worthless gift to the other. This is in fact
a reinterpretation of the example taken from Van Damme [1989] where there is
an opportunity to burn a certain amount of money. By forward induction, Van
Damme shows that a gift is made with positive probability. The gift serves as a
credible threat of playing a particular strategy.
Generally, in accordance with empirical data, gifts as signals should be inadequate
or need at least not necessarily be adequate. For example in the games by Camerer
[1988] and Kranton [1996b] inadequacy is a prerequisite. In the model of Van
Damme [1989] inadequate as well as adequate gifts are possible16 . Reciprocity
can also be explained by most signalling models. This is a strong point of the
signalling approach. An obvious restriction of this approach is that it is related
to informational asymmetries. This is likely to be most relevant for relationships
16 This
is simple to prove. The argument in Van Damme [1989] (see in particular his fig. 5) is independent of
what the other player gets. The strategy of each player takes the other player’s strategy as given and as a result
the adequacy of the gift does not matter.
51
Discussion
that are short in nature or are in their beginning phase. Business gifts may belong
to this category, Christmas gifts less so.
2.4 Discussion
The main focus of this chapter has not been to provide a unique unified theory of
gift-giving. Rather, it aims at exposing competing theories and to evaluate them
on their explanation power of accounting for the two characteristics reciprocity
and adequacy. The possibility of coexisting motivations for gift-giving should not
be disregarded. Perhaps the motivations differ between different kinds of gifts,
different people, or different time periods.
The chapter did, however, shed some light on which motivations to look for
in a wide range of situations. Gifts with the purpose of exchange can only be
expected when players have a sufficiently long horizon and where the players are
familiar to each other. Altruism cannot explain inadequate gifts, although it is
likely to partly explain charity. Social approval can to a certain extent explain
charity, although the scope is somewhat limited: at least some people should
get to know about the act. Fairness may explain charity as well, and describes
many experimental games relatively well. Signalling explanations are powerful to
explain short-run or beginning relationships. They seem not applicable to long
run relationships, where information asymmetries have dissolved over time.
I would also like to argue for a hybrid model of gift-giving: one where different
motivations act and interact simultaneously. Emotions and strategical actions
may both play a role. What I have in mind here is for example that people are
not intrinsically fair but still like to appear to other people as fair, and accomplish
that by giving a signal that one is fair. This combines the different motivations
fairness, a need for approval, and signalling. This particular example comes also
out of the experiment by Güth and Van Damme [1998], where a third dummy
player is added to the ultimatum game (see section 2.3.3). Recall that one of
the main results is that only marginal offers are proposed to the dummy player.
They conjecture that the proposer was not intrinsically motivated by fairness
considerations but that they do not want to be fair, but rather want to appear
fair, e.g. to prevent a rejection by the responder.
Knowing the motivations helps to design efficient incentive systems. There is
one classical example where misinterpreting the motivations to give led to a wors-
Chapter 2. The Economics of the Gift
52
ening of the situation. When Titmuss [1970] examined the blood market, he found
that when it was tried to stimulate blood donations to reduce the shortage of
blood, by giving monetary compensations, this resulted in less blood donations
rather than more. If altruism were the only explanation of blood donation (which
is not very likely, see section 2.3.4) the decision to compensate would be understandable. If other factors play a role, such as a need for approval, the effects of
monetary compensations on these motivations should not be disregarded. Indeed,
in chapter 3 I show that monetary compensations may decrease the willingness
to give by obstructing possible approval (less approval is rewarded if the sacrifice
is less).
The blood market is not the only example. In another example taken from
Gneezy and Rustichini [2000] the other way around is also found, namely that the
parents that arrived too late to collect their children at day-care centers increased
in number after a fine was imposed (the gift is here the additional time that was
spent by the employees). In the interpretation of opening markets, the creation
of a market is in this case the cause of crowding out of the gift, something that is
indeed often observed (see for instance Frey [1997a,b], and Frey and Jegen [2002]
but also the discussion in Arrow [1972]). This conclusion is similar in spirit to
that of Holländer who concludes that it may well be the case that ”the opening of
a market (...) reduces voluntary contributions” (Holländer [1990, 1165]). I refer to
section 3.2.4 for a more detailed discussion about the interaction between markets
and gifts. As one can see already, the appropriate incentive mechanism takes into
account other motivations than selfishness. A name in the records may have more
impact than a reward.
A final note on where all these motivations come from. Carmichael and MacLeod
[1997] have shown that gifts as a signal at the beginning of a relationship can lead
to cooperation. They also showed that this institution is evolutionarily stable. The
intuition is as follows. A gift is a signal of trustworthiness and has to be given at
the beginning of any new relationship. If both agents in a partnership conform
to this custom they only have to give once and stay in the partnership forever. A
free rider, on the other hand, is detected after one period and has to search for
a new relationship again. This cannot be a strategy that often pays off, since for
every new relationship a gift has to be incurred.
Other authors have shown that many other emotions such as anger and altruism can also be evolutionarily stable (see e.g. Güth and Kliemt [1994], [1998]).
53
Discussion
Emotions like anger may result in lower current payoffs, since punishing is usually
also costly to the punisher, but may still induce behavior that is evolutionarily
successful. Defectors are deterred away by the threat of punishments. This explains such sentiments as having emerged from selection pressures, which has
resulted in modern institutions such as gift-giving.
3
The Demand for Social Approval as a
Motivation to Give
The Kwakiutl were once one of the major tribes of the Norhwest Coast. They
were quite wealthy, often even richer than many of the other settlers in that
area (see Codere [1950]). They have intrigued anthropologists not in the least
because of their ceremonies. The best known ceremony is the so called potlatch,
a gift-giving ceremony. During the potlatches, many blankets and large amounts
of copper were being given away to other tribesmen. However, this was done in
a way that seems particularly inefficient: many of the gifts are destroyed on the
spot. The anthropologist Mauss observed that during these potlatches ”they go as
far as the purely sumptuary destruction of wealth” (Mauss [1926, 6], witness also
the following quote:
(...) she ordered one of her kinsmen to tow it [copper] to sea behind a
canoe and to cut it adrift in deep water and let it sink. ”This is my
gift to you, O chief.” (Drucker and Heizer [1967, 105]).
Although less visible and certainly less extreme, the same behaviour can also be
found in gift exchanges that take place in modern societies. Rather than giving
0A
shorter version of this chapter appeared in the Journal of Institutional and Theoretical Economics, 2002
vol. 158 (3), 464-482 under the title: The Demand for Social Approval and Status as a Motivation to Give. I am
indebted to Michèle Belot, Jeffrey James, Luuk van Kempen, Fieke van der Lecq, Sjak Smulders, participants
at the European Economic Association Conference (Lausanne, 2001), and especially to Theo van de Klundert
and two anonymous referees for very valuable comments and suggestions on earlier versions.
Chapter 3. The Demand for Social Approval as a Motivation to Give
56
in cash, gifts are mostly in kind, thereby destroying some of the monetary value
to the receiver. This chapter tries to shed some light on the motivations behind
such, and other related behaviour.
3.1 Introduction
Gift-giving has mainly interested anthropologists because it has been taken as a
primitive mode of exchange. Relying both on a double coincidence of wants and on
the existence of trust between the agents, gift-giving clearly seems to be inferior to
the market mechanism. Yet, gift-giving is still widely observed even in countries
with well-developed markets. Indeed, gift-giving sometimes even flourished in the
presence of a market economy (Gregory [1989]).
The difficulty is not to explain gift-giving per se. One can simply attribute
utility to the act of giving, or, in the terminology of Andreoni [1990], a warm
glow feeling. Rather, there are some stylized facts related to gift-giving that have
to be explained but are puzzling from a standard economics point of view. These
stylized facts include reciprocity, inadequate giving, and a negative correlation
between monetary compensations for gifts and the level of gift-giving. Reciprocity
refers to the observation that, although voluntary on guise, gifts appear in fact to
have strong reciprocal properties. Inadequacy points to the fact that gifts should
be in cash in order to maximally satisfy the preferences of the receiver, but often
they are not. Finally, it is observed that gift-giving is sometimes reduced after
compensation is offered. Neither one of these stylized facts can easily be explained
by standard economic arguments.
The main point to this chapter is to explain gift-giving by means of a demand
for social approval and status; two factors deeply rooted in human nature. The
idea that approval motivates gift-giving is in my opinion intuitively an appealing
hypothesis. Public goods experiments show that familiarity with the identity and
actions of other players leads to significantly higher contributions (Andreoni and
Petrie [2000], Gächter and Fehr [1999]). Earlier Satow [1975] has shown that the
increase in donations in public conditions as compared to private conditions is
strongly correlated to the individual’s need for approval.
Introduction
57
Although not entirely neglected in the economics literature,1 the concepts of
social approval and status have, above all, found an eminent place in sociology
and social psychology. It plays the same part there as money does in economics:
recognition by others is regarded as a primary source of satisfaction (see for
example Homans [1961]; Kenrick, Neuberg, and Cialdini [1999]; and Coleman
[1990]). Social approval in this sense appears to be a functional substitute for
money. Hence, a transaction that is unequal in monetary terms — as a gift is —
can in principle still be in balance as long as approval is awarded.
Gift-giving indeed appears to be a virtue and a source of prestige (Polanyi
[1957], Schwartz [1967]). The way in which approval is obtained is however complex. Evidently, approval will be higher the more the gift is valued. But it also
turns out to depend on the sacrifice incurred by the giver. Moreover, approval
seems to be closely linked with status. Recognition and status are often mentioned
in one and the same breath (for example in Schwartz [1967, 7], and Harsanyi
[1969, 523]). Taking these factors into account, the stylized facts become natural
implications of the model.
Other implications of taking into account the taste for social approval follow
from relating gift-giving to the market institution. Since the market is in its purest
form an anonymous institution, no social approval is obtained in a market exchange. On the other hand creates the market incentives to maximize adequacy.
Gift-giving as an exchange mechanism does allow for acquiring approval but in
general fails to maximize adequacy. Hence, in choosing between a gift-exchange
or a market exchange the trade-off to be made is that between approval and adequacy. However, spontaneous order does not necessarily lead to the most efficient
institution because the links between the institutions are shot through with externalities (Dasgupta [2000]). Unfortunately, trying to correct for any inefficiencies
with a standard economic tool like money compensations does not always resolve
this and may even have the opposite effect of worsening the situation. This is in
line with what the model is able to predict and with ample empirical evidence.
Roughly, the basic line of argument runs as follows. In the model, two individuals are playing a sequential move gift-giving game. Each of them has preferences
for both a consumption good and social approval. Player 1, then, makes a gift
for which approval is awarded. But lagging behind in the status race for wanting
1 See
Akerlof [1997], Holländer [1990], and the references therein.
Chapter 3. The Demand for Social Approval as a Motivation to Give
58
more approval than the other, player 2 finds it profitable to react by making a
countergift. This explains the so often found reciprocal behavior. Inadequacy is
just as easily explained. In trying to prevent player 2 from catching up in terms of
status, player 1 can deliberately devalue his gift by giving in kind rather than in
cash. This makes it more expensive for player 2 to reciprocate. Finally, note that
a compensation reduces the sacrifice needed to make a gift. This is positive in
terms of consumption, but it is also likely to reduce the awarded social approval
for the gift. Whereas with standard assumptions on preferences a compensation
should increase gift-giving, with a preference for approval it is ambiguous whether
compensation increases or reduces gift-giving in equilibrium.
The setup is as follows. The subsequent section is concerned with deriving
the basic properties of gift-giving from a simple model. Building on empirical as
well as experimental evidence, an extensive account is given on what the social
approval function should look like and what kind of consequences it has for the
way economists think about the workings of the market institution. Section 3.3
relates the model to the existing literature. Finally, section 3.4 concludes.
3.2 Social approval, status, and gift-Giving
“Gratitude is bestowed on a giver.” Aristotle, Nichomachean Ethics.
3.2.1 The basic model
In this section, the tentative explanations of the stylized facts are made more
precise. The model is highly stylized and is only meant to be suggestive in explaining how social approval can affect decisions. I believe however that the key
insights are not sensitive to the specification of the model and that they would
survive in a more general framework.
In the model there are two players. Player 1 is the fist mover in a sequential giftgiving game, and player 2 follows. Their decision variable is the amount of time
spent doing volunteer work (liv ). The rest of their available unit of time (1 − liv )
is devoted to working in the market sector at wage wim . With the income that is
earned on the market, wim (1 − liv ), the consumption good x can be purchased,
which is available at unity price.
The time spent doing volunteer work is a gift and contributes to the other
person’s consumption level. For simplicity, it is assumed that the gift is the same
Social approval, status, and gift-Giving
59
consumption good x. The gift increases the other person’s consumption by the
amount of δi liv . This way, the parameter δ can be interpreted as a measure of
adequacy. By adequacy is meant how the receiver’s utility of consumption is
increased relative to the costs incurred by the giver. A more adequate gift increases
the receiver’s utility of consumption more given the costs incurred by the giver.
Note that standard microeconomics arguments tell us that it can never be worse
in terms of utility to get a gift in cash rather than in kind. As a result, a cash-gift
is in general more adequate than a gift in kind. This can be translated back into
the model as follows. If δ i = wim then the gift is exactly identical to giving money.
Hence, this case is interpreted as if it were a cash-gift. If δ i < wim , the gift is worth
less than the cash-equivalent. The latter case is interpreted as a gift in kind.2
In sum, total consumption of good x by player i is given by:
xi = wim (1 − liv ) + δ j ljv .
(3.1)
If utility is only derived from consumption, as is usually assumed in economics,
neither one of the players will give. Whatever player 1 gives to player 2, it is
optimal for player 2 not to make a countergift. Foreseeing this behavior, player 1
does not make a gift as well.
The main departure from standard models is the inclusion of a preference
for social approval as well as for status. There are good reasons to do so. The
worry about status and the ’thirst for approval’ are among the most recurring
themes in anthropology (Wright [1994]). They thus seem to be deeply rooted in
human nature. The existence of suchlike sentiments possibly has emerged from
selection pressures because emotive motivations indirectly induce behavior that
is evolutionary successful (see Güth and Kliemt [1994], [1998]). Each of these
emotions is explained in more detail below.
Social Approval
The first building block is to take into account the taste for approval. It is beyond
any doubt that a preferences for approval exists. Man is a ’social being’ whose
economy is submerged in his social relationships. Polanyi [1957, 46] considers this
δi > wim , a gift is worth more to the receiver than its cash-equivalent. As said, this is not the standard
case, but one can think of examples where it is a possibility, for instance when the receiver has incomplete
knowledge over his own preferences.
2 If
Chapter 3. The Demand for Social Approval as a Motivation to Give
60
to be the “one conclusion [that] stands out more clearly than another from the
recent study of early societies.”3 Indeed, the preference for social approval may
very well be as important as the preference for consumption goods (Harsanyi
[1969], Sugden [1989], Dasgupta [2000]). But it seems that the actual obtaining of
approval is a rather complex process. In order not to complicate matters any more
than is necessary for current purposes, the focus is on two elementary properties.
First, the higher the value of the gift as judged by the receiver, the more the
gift is approved. The social approval function for player i should consequently be
increasing in δ i .4 Second, approval is increasing in the sacrifice made by the giver.
Sacrifice is likely to be something relative to what you earn. A gift of one dollar
by the poor is approved more than the same donation by a millionaire. This is
exactly the behavior that Pruitt [1968] finds. In his experiment, more reward was
provided by the receiver if the giver had sent out 80% of his endowment of $1
than if he had sent out 20% of his endowment of $4, presumably because the
sacrifice is larger in the former case. The simplest measure of sacrifice is the wage
level. The higher the wage, the more consumption is forgone in order to give. So,
a higher wage corresponds to a higher sacrifice. This amounts to a social approval
function that is increasing in wim . Further empirical evidence for this is found by
Robben and Verhallen [1994].
Status — Why humans get ulcers
“Men have an immoderate love of pleasure, influence, prestige, power
— in a word, wealth.” F. Bastiat, Economic Sophisms.
It is conventional in economics to assume that a higher income generates a higher
well-being. When psychologists try to measure the happiness of people they indeed find such a relationship on the individual level. Paradoxically, however, the
average satisfaction level is remarkably stable over time, despite significant increases in per capita income (see for example Frank [1997]).
It seems that the disappointing increase in happiness is closely related to a
question that occupied Hume more than two centuries ago: How can it be that
3 For
a similar account in modern sociology see Coleman [1990].
seems evident but may in fact not be entirely trivial. In some cases a very large gift may actually
cause embarrassment thereby decreasing approval, as in the case one would get diamonds on a first date. I do
not pursue this point any further here.
4 This
Social approval, status, and gift-Giving
61
our impressions of the same object can at one time be an admiration of its bulk,
and at another to despise it for its littleness? Hume found the solution in a careful
examination of human passions:
”So little are men governed by reason in their sentiments and opinions,
that they always judge more of objects by comparison than from their
intrinsic worth and value” (Hume [1886, 158]).
Comparison is the keyword here. Much like a certain mountain looks big next
to a smaller mountain yet only little next to a bigger one, the same amount of
wealth is pleasurable if you have more of it than your neighbours, but becomes
frustrating when you have less. In other words, the apparent solution to the
paradox in question can be found in the supposition that people deeply care
about status. That is, they certainly do care about wealth, but only insofar as
it increases their prestige, a form of status. This role of wealth, also present in
the quote by Bastiat at the outset of this section, implies that the increased wellbeing of an individual from a higher income dissolves once the other people in her
reference group reach the same level of prestige that she was privileged to enjoy
before. A general increase of the wealth of a nation has thus no profound effect
on the general happiness of people.
The idea that people care about status has often been recognized before in
the economic literature5 . The reason for that is clear, as already as far back as
Veblen [1899/1953, 80] it is acknowledged that in many articles ”... the traces of
conspicuous waste, or at least the habit of ostentation, usually become evident on
a close scrutinity”. Darwinian anthropologists have put forward the hypothesis
that worries about status is one of the most recurring patterns in all cultures
(see Wright [1994]). Furthermore, research in biology confirms that up to the
present day humans are equipped with a hard-wired preference for status. The
achievement of status involves physiological consequences. For example, it turns
out that more of the neurotransmitter serotonin (which is, as it happens, also
used in many antidepressants) is being secreted when humans (or animals) are in
the position of a leader (see Wright [1994] and Frank [1985] for a more elaborate
5 For
an early contribution that employs status as an assumption: see Leibenstein [1950]. More recent
contributions include Cooper et al. [2001] and Corneo and Jeanne [1997]. Van Kempen [2003] points out that
even poor people have status concerns. See also the references therein and in the main text of this chapter for
more contributions.
Chapter 3. The Demand for Social Approval as a Motivation to Give
62
treatment). In his Why zebras don’t get ulcers, Sapolsky [1998] even goes as far
as associating sudden cardiac deaths (and human ulcers, for that matter) with
major stressors such as the loss of status, and cites several studies that are indeed
congruent with this view. Evolutionarily speaking, all this makes sense: a high
level of status gives better access to resources, such as food, and increases the
probability of successful reproduction as it attracts partners (Frank [1985], Wright
[1994]).
Much of the research done has focused on domination-oriented status, and
mostly among animals. But, according to Barkow [1989], as a consequence of
selection contemporary human prestige has a more symbolic nature than the
agonistic dominance found among primates. Prestigious objects or acts signal an
increased ability and willingness to make paternal investments in offspring.
Giving can be a source of status as well, in the form of prestige. This is for
example the case in the consumption domain. Wealth increases prestige. When
wealth is not so visible to outsiders, gifts can enhance status by giving a signal
about the wealthiness of a particular person.6 I, however, follow Holländer [1990]
in employing status effects in the approval domain. Holländer [1990] cites several
studies in which it is argued that people not only want to be admired, but also
want to be more admired than others. This is also experimentally verified by
Gächter and Fehr [1996]. They find that social approval decreases in the average
level of contributions by other subjects and point to the similarity with status
effects. Implicitly, Forge [1972] finds this also to be the case in the more archaic
societies in New Guinea when he remarks that a gift exchange is never perfectly
balanced so that if the recipient fails to outdo the giver, it is the giver who gains
and the receiver who loses. Or, in the more explicit words of Gregory [1989, 110]
”... gift exchange necessarily introduces status inequalities”.7
Preferences
Based on these building blocks, the following measure for approval is proposed:
si = β i liv − αβ j ljv ,
6 See
α ≥ 0.
(3.2)
Glazer and Konrad [1996] for a formal model.
how many others will not be as tempted as Ng is to give as expensive gifts as the schoolmates of one’s
child receive? See Ng [1997].
7 And
63
Social approval, status, and gift-Giving
Every unit of voluntary labor is weighted by the function β i = β(δ i , w̃i ), which
is increasing in both of its arguments: the weight is larger if the value of gift is
higher (δ i ) and if the sacrifice (w̃i ) is greater, as measured by the forgone wage
of the giver (wim ). The parameter α in equation (3.2) reflects the degree of status
orientation.
Finally, preferences are represented by a utility function that is for simplicity
additive in consumption and net social approval:
ui = ux (xi ) + us (si ).
(3.3)
It satisfies the usual assumptions with respect to x and s: u0x , u0s > 0 and u00x ,
u00s ≤ 0, where (double) primes denote first (second) derivatives. At this point it
should be noted that in practice the degree of status orientation is not uniform
among people (see Wright [1994]) and neither is the intensity of the need for
approval (Satow [1975]). Including such heterogeneity would not alter the basic
arguments of this chapter and is for that reason left for future research.
Equation (3.3) is a possible representation of the theory of social behavior that
Homans has in mind when he speaks of ’social behavior as exchange’ (Homans
[1958, 606]). In his view, social behavior is an exchange of goods, including nonmaterial ones such as approval. Now note that although an exchange of material
goods can be accomplished in any social setting, the accomplishment of approval
is tied to social interaction. That the utility derived from a good depends on
the social context is a simple extension of Lancaster’s theory of consumption
(Lancaster [1966], Hirsch [1976, 85]). In the view of Lancaster [1966], goods as such
are not the direct object of utility. Instead, utility is derived from characteristics
and goods are bundles of characteristics. The environment of exchange can be
seen as one of those characteristics. Compare also Bowles [1998, 87] who notes
that “the terms on which [people] are willing to transact depends on the perceived
relationship.”
This amounts to a useful interpretation of equation (3.3) in that it captures
various exchange regimes characterized by different intensities of social interaction, with the strictly anonymous market exchange as the extreme where no social
interaction exists at all (us = 0, although practically speaking even the market
is characterized by some social interaction rather than by complete anonymity).
Hence, in studying market trade there are good reasons to neglect the role of
any such sentiments since they are ruled out by anonymity. But things are dif-
Chapter 3. The Demand for Social Approval as a Motivation to Give
64
ferent when studying gift-giving where, contrary to abstract markets, trade is
not anonymous. Gift exchange is above all a social relationship (Gregory [1989]).
In such a case the taste for approval strongly influences behavior. Suggestive in
this respect is the recent finding by Gächter and Fehr [1999] that in experimental
settings some minimal social familiarity generates a significant rise in cooperation.
3.2.2 Sequential move equilibrium
The game is solved by backward induction. The problem of player 2 is to solve
for:
l2v∗ ∈ arg max
u2 (l1v , l2v ),
(3.4)
v
l2
l1v .
for any given
Ignoring parameters, this gives a reaction function of player 2 of
v∗
the form l2 = f2 (l1v ). Player 1 takes this behavior of player 2 into account and
therefore solves for:
l1v∗ ∈ arg max
u1 (l1v , f2 (l1v )).
(3.5)
v
l1
The solutions to (3.4) and (3.5) constitute an equilibrium. Throughout the focus
is on an interior solution since this is the only interesting case.
Reciprocity
The first result is obtained from the properties of player 2’s response function
f2 (l1v ). Since this function relates the optimal gift of player 2 to the gift of player 1,
it predicts whether reciprocal behavior should be observed or not. Note first that,
as argued earlier, without a preference for approval (us = 0) the optimal response
of player 2 is not to make a gift, independent of the gift by player 1. Without a
preference for approval, the only effect of a gift would be a loss of consumption.
Player 1 foresees that f2 = 0, so that it is also for him optimal not to give. No
gift-giving occurs even if there are mutual gains of giving (δ 1 > w2m and δ 2 > w1m ).
The cause of this is that player 1 has to rely on a countergift by player 2, but
he has no reason to trust player 2 on this. An explicit or implicit contract can
of course solve this problem of trust. But reciprocal gift-giving is also observed
in situations where no explicit contract exists, or at most an incomplete one,
and where an implicit contract is not credible. Mauss [1925/1980], for example,
portrays gift-giving among tribes, and finds that reciprocity is one of the basic
elements. And Akerlof [1982] gives a description of labor contracts, where wages
Social approval, status, and gift-Giving
65
are set above the minimum acceptable standard and these are reciprocated in the
form of higher efforts. These results of reciprocal gift-giving without complete
explicit contracts are replicated in many laboratory experiments (see for example
Fehr and Schmidt [1999] and Bolton and Ockenfels [2000]). In many cases these
were one-shot games so that reputation cannot enforce credible implicit contracts
either (see for example Cooper et al. [1996], and Gächter and Falk [1999]). The
point of all this, is that apparently contracts are not always necessary to induce
reciprocal gift-giving. As it turns out, a preference for social approval is enough.
That gifts can be positive in equilibrium is quite obvious: all one needs to
assume is that the gain in net social approval outweighs the loss of consumption
in utility terms. More interesting is to account for reciprocity. This is established
in the following proposition:
Proposition 1 (Reciprocity): With a preference for social approval, the optimal
gift of player 2 is positively related to (i) the gift of player 1 (∂f2 /∂l1v > 0), and
(ii) the degree of status orientation ( ∂f2 /∂α > 0).
Proof. By definition of f2 :
respect to l1v gives:
∂ 2 u2 (·)
∂lv2 ∂lv1
+
∂u2 (l1v ,f2 (lv1 ,·),·)
∂lv2
∂ 2 u2 (·) ∂f2 (l1v ,·)
· ∂lv
∂l2v2
1
≡ 0. Differentiating both sides with
¯
∂f2 (l1v )
∂ 2 u2 (·)/∂l2v ∂l1v ¯
= 0 ⇔ ∂lv = − ∂ 2 u2 (·)/∂lv2 ¯ v∗ . By
1
2
l2
the second order condition of maximization, the denominator of the right hand
2
∂f (lv )
2 (·)
side is negative, so that sign ∂l2 v1 = sign ∂∂luv ∂l
v = sign
1
2 1
h
i¯
2 ∂x2
2 ∂s2 ¯
u00x ∂x
+ u00s ∂s
. The first term in brackets is strictly positive. The
∂lv ∂lv
∂lv ∂lv ¯ v∗
2
1
2
1
l2
at the optimum sign
the second part.
∂f2 (l1v )
∂l1v
∂s2
∂l1v
= 0, and positive if α > 0. Hence,
¯
∂ 2 u2 ¯
2
> 0. Similarly, sign ∂f
=
sign
> 0, proving
∂α
∂l2 ∂α ¯ v∗
second term is zero if α = 0, in which case
l2
If the players care about approval, they behave in a reciprocal manner. This
result in itself (i.e. (i) in Proposition 1) holds in principle for every degree of
status orientation (including α = 0), but is strengthened by a higher degree of
status orientation (see (ii)). This partly solves the trust problem. Player 1 has
now reason to believe that player 2 makes a countergift, simply because it is
in player 2’s own benefit to do so. This does not mean that player 2 always
reciprocates the gift of player 1, just like an altruist not always gives something.
What it means is that the more player 1 gives, the more he is likely to get
something back. The intuition behind this result is the following. The gift of
player 1 increases the consumption level of player 2, thereby decreasing marginal
Chapter 3. The Demand for Social Approval as a Motivation to Give
66
utility of consumption. Insofar as status plays a role, the gift also decreases the
net social approval received by player 2, thereby increasing the marginal utility of
net social approval. Because a countergift has opposite effects on the consumption
and the net approval level, a gift by player 1 creates incentives for player 2 to
give as well.
The foregoing has concentrated on reciprocity in the sense of rewarding generous gifts with countergifts, also called positive reciprocity. The other side to the
coin is negative reciprocity, by which it is commonly understood that players who
are stingy are punished for being so. For instance, low offers in ultimatum games
are often rejected, making both players worse off.
The terms ’rewarding’ and ’punishing’ seem to imply a (fairness) norm to which
players are expected to comply to. Positive deviations are rewarded, negative
ones punished. Rejections in anonymously played ultimatum games and other
public good games with the option to punish is indeed best explained by feelings
of resentment, negative reciprocity in other words (Fehr and Gächter [2000]).
This has not much to do with approval. Nonetheless does a demand for social
approval and status give some hints as to what will happen in those games when
there would be some familiarity among the players. First, players would have
the opportunity to get approval for punishing stingy players, strengthening the
incentives for punishments. Second, apart from feelings of resentment, players
might be less willing to accept low offers in the ultimatum game for fear of a loss
of status. Since player 2 can either accept (A) or reject (R) in the ultimatum
game, his gain in utility from accepting the gift of player 1 is8 :
∆u2 = u2 (A) − u2 (R) ≈ δ 1 l1v u0x − αβ 2 l1v u0s,
(3.6)
which is the sum of the gain in consumption and the loss in net approval. To accept
but not being able to reciprocate means a sure loss in net approval, hence the
observation that “Charity is still wounding for him who has accepted it” (Mauss
[1925/1980, 65]). For high levels of status orientation this status loss outweighs
the consumption gain and makes rejection sensible. Note that the receiver would
have preferred to give back (f2 > 0) but is by construction forced to l2v = 0 (with
l1v > 0), or to refrain from any gifts at all (both l1v = l2v = 0).
derive equation (3.6), note that ∆u2 /∆l1v ≈ ∂u2 /∂l1v = δ1 u0x − αβ 1 u0s and ∆l1v = l1v − 0 since the
reference point is no gifts.
8 To
Social approval, status, and gift-Giving
67
Interestingly, and perhaps related, Mauss [1925] also points out that the root
of the word gift is the same as that of poison. Harm befalls on the recipient of
a gift, as he is challenged to catch up in the competitive race for status. This
imbalance that is created by a one-sided gift is also nicely captured in the Indian
verb quoted in Brigham [1991, 300]: ”Why do you hate me? I’ve never even helped
you”.
A cautious remark should be made here. A rejection itself can be perfectly
rationalized within the current model. But it turns out that while low offers
are sometimes rejected, high offers are normally accepted. It is easily checked
that with u00s < 0 this behavior is inconsistent. This indicates that the approval
function is more complex than the one operated, for instance, one that is convex
above a reference point and concave below. Although most of the results would
go through under this extension, it leaves open the question how such a reference
point is determined, and is therefore left for further research.
One would thus expect that rejections are more often observed in ultimatum games where there is familiarity among the players. Translated to out-oflaboratory situations, one would expect a needy person to more easily accept
a gift from an anonymous source than from a person he can identify, since the
latter introduces a loss in net social approval. Finally, one would expect that if
possible, players would rather want to give something back. At least for the latter statement there is already some evidence (see Kenrick, Neuberg, and Cialdini
[1999]) but it would certainly be worth to investigate these and related issues
more thoroughly.
Adequacy
The presence of a taste for approval secures that a gift is usually reciprocated.
This allows an exchange to take place without the support of contracts. But
in the current gift-giving game, the exchange is only an intermediate variable
with approval as the ultimate objective. It is interesting to see how this role of
exchange affects the properties of the gift. This relates to a puzzle that concerns
the adequacy of gifts. A cash-gift is more adequate than a gift in kind. But on
many occasions, it is quite unusual to give cash. Research by Webley and Wilson
[1989] is illustrative of this. Their questionnaires reveal that subjects prefer to
give presents rather than money. Likewise, Caplow [1982] finds that in the data
he collected less than 9% of the Christmas gifts were in cash.
Chapter 3. The Demand for Social Approval as a Motivation to Give
68
This behaviour is puzzling at first sight because of the assumption that gifts
serve to gain approval and presumably less adequate gifts are less approved. But
on closer inspection, there is a rationale behind this behavior. Player 1 likes
getting a countergift in that it increases his consumption. But he dislikes it in
that it decreases his net approval. Clearly, if the gift of player 2 would be a little
bit lower, he can be either better or worse off, depending on the relative decrease
in consumption as compared to the relative increase in net social approval. This
means that there are incentives for player 1 to manipulate the gift of player 2.
Suppose he indeed wants to lower the gift of player 2. According to the following
lemma, he can do this by decreasing the adequacy of his own gift.
Lemma 1 : Player 2 reciprocates more if he receives more adequate gifts, that
is, ∂f2 /∂δ 1 > 0.
Proof. With the introduction of the second decision variable for player 1, δ 1 , the
optimal gift of player 2 is dependent both on l1v and δ 1 : l2v∗ = f2 (l1v , δ 1 ) (ignoring
∂u (lv ,f2 (l1v ,δ1 ),δ1 )
parameters). By definition of f2 : 2 1 ∂f
≡ 0. Differentiating both sides
2
¯
∂ 2 u2 /∂l2 ∂δ1 ¯
2
with respect to δ 1 gives, after rearranging: ∂f
=
−
. By the second
∂δ1
∂ 2 u2 /∂l2v2 ¯lv∗
2
¯
∂f2
∂ 2 u2 ¯
order condition the denominator is negative, so that sign ∂δ1 = sign ∂lv ∂δ1 ¯ v∗ > 0.
2
l2
The intuition behind this intermediate result is that by lowering the adequacy
of his gift, player 1 contributes less to the consumption level of player 2, keeping
the marginal utility of consumption for him at a high level. This makes it more
expensive for player 2 to give. Besides this effect, it also increases the net status of
player 2, decreasing the marginal utility of social approval. Both effects reduce the
incentives for player 2 to make a gift. The behaviour by the kwakiutl of destroying
gifts (see the introduction to this chapter) can also be understood along to these
lines.
To derive the desired result formally, I say that the gesture is relatively important if the approval rate is not very sensitive to the value of the gift: i.e. ∂β i /∂δ i
is low. In this case, it is the act of giving that counts, and not so much the gift itself. This is not so unreasonable. As Holländer [1990, 1161] puts it: “we generally
approve of cooperative behavior even if it does not make us significantly better
off.”
With the foregoing definitions, I obtain the following result:
Social approval, status, and gift-Giving
69
Proposition 2 (Adequacy): With a preference for social approval, player 1 has
an incentive to reduce the adequacy of his gift (du1 /dδ 1 < 0) if the degree of
status-orientation (α )is sufficiently high and the gesture is relatively important
(∂β 1 /∂δ 1 low).
v
v
Proof. Player 1’s utility
is given by:
h
i u1 =h u1 (l1 , f2 (l1 , δi1 ), δ 1 ). The total derivative
1
1 ∂f2
1
1 ∂f2
is given by: du1 = ∂u
+ ∂u
dl1 + ∂u
+ ∂u
dδ 1 . The term in the first
∂l1v
∂f2 ∂l1v
∂δ1
∂f2 ∂δ1
h
i¯
∂u1
∂u1 ∂f2 ¯
1
brackets is zero by the first order condition. Hence: du
=
+
. The
dδ1
∂δ1
∂f2 ∂δ1 ¯ v∗
l1
first term in brackets is the direct effect and is positive ∂u1 /∂δ 1 = (∂β 1 /∂δ 1 )l1 u0s:
a higher δ 1 increases net approval at the rate (∂β 1 /∂δ 1 )l1 u0s. This term vanishes
as ∂β 1 /∂δ 1 becomes smaller. The second term in brackets in the indirect effect
due to the response of player 2. Since by Lemma 1 ∂f2 /∂δ 1 > 0, the sign of the
1
indirect effect is given by the sign of ∂u
= δ 2 u0x − αβ 2 u0s . Using the FOC and the
∂f2
additional assumption that w1m > δ 2 f20 it is easy to show that the sign is negative
for α > (β 1 /β 2 ) · (δ 2 /w1m ). Hence, the total effect is negative if ∂β 1 /∂δ 1 is low and
α is sufficiently large. The assumption that w1m > δ 2 f20 merely states that gifts
require a sacrifice and are made to gain approval, not to increase consumption.
This seems reasonable when one speaks of social approval as a motivation to give.
Let me reflect for one moment on the results so far. In the introduction to this
chapter, I have described the gift-giving ceremony of the Kwakiutl: the Potlatch.
Here, I would like to argue that the need for social approval and status may be
an accurate description of their gift-giving practices. Many of Mauss’ observations which often clearly point to status concerns, including his observations that
”Face is lost forever if [a return gift] is not made” (Mauss [1926, 41] and that
large gifts are made by men ”in order to outdo their rival” (Mauss [1926, 6]9 .
Also congruent with a demand for social approval and status is his mentioning
of the fact that gifts are often ostentatious. Interestingly, we also saw that the
Potlatch is characterized by an extreme degree of inadequate gift-giving, in line
with proposition 2. According to Godelier [1996, 56], the goal of this is explicitly
to ”make it difficult or impossible to give back the equivalent”. The broken copper is seen as a victory much like as a killing of a rival would have been. It is a
9 It
is in this respect also interesting to note that Mauss [1926] traces the origin of the word "gift" back and
finds that is has the same roots as poison.
Chapter 3. The Demand for Social Approval as a Motivation to Give
70
fight for prestige, a war that is fought with property as weapons, to paraphrase
Codere [1950]. The last thing you want is to provide your enemy with weapons.
The destruction of copper does exactly prevent that.
Efficiency
The incentive to give inadequate gifts has warranted researchers to conclude
that gift-giving creates a deadweight loss. Subsequent calculations accredited to
Christmas gifts a deadweight loss in the range of 10-30% of their value (Waldfogel
[1993]). Yet, one cannot conclude from inadequacy itself that gift-giving creates
an inefficiency, in the sense that everybody could be made better off. In the research done by Waldfogel [1993], and also in the follow-up studies by Solnick and
Hemenway [1996] and Waldfogel [1996], it is tacitly assumed that the giver is indifferent between giving in cash and in kind. However, as found in questionnaires
(see e.g. Webley and Wilson [1989]) and in line with Proposition 2, givers are not
at all indifferent but clearly prefer to give in kind. Hence, inadequacy does not
imply inefficiency. However, the following proposition nevertheless shows that in
equilibrium gift-giving is inefficient.
Proposition 3 (Efficiency): There exists a feasible distribution of endowments
that is a Pareto-improvement upon the gift-giving equilibrium.
Proof. The proof follows in a fairly straightforward way from proposition 1.
Note that the reaction function f2 of player 2 is an upward sloping curve in the
(l2v , l1v ) plane. Recall that this curve is the set of points where player 2’s utility is
maximized for any given l1v . Thus, the line f2 cuts each of player 2s indifference
curve at its maximum. Hence, the indifference curve has slope zero everywhere
along f2 . Player 1 chooses a point on this curve that maximizes his utility. This
is necessarily a point where the slope of his indifference curve is tangent to the
reaction function of player 2. Hence, in equilibrium the slope of the indifference
curve is positive for player 1. Since the slopes of the indifference curves differ
between the players, there exists a point where both could be made better off.
The cause of this inefficiency is twofold. First, the preference for status introduces
an inefficiency. This is clear: both players put resources in the race for status, but
only the net result counts. This is most clearly seen for the case where α = 1.
Would both players spend the same amount on gifts, none of them would derive
71
Social approval, status, and gift-Giving
utility from it with regard to social approval, since net social approval would equal
zero. Second, player 1 acts strategically by taking the optimal response of player
2 into account. He does not, however, take into account the effects on player 2s
welfare. To derive the social optimum, he would have to take into account the
welfare of both players.
3.2.3 Simultaneous move equilibrium
So far, the focus has been on the sequential move gift-giving game. The reason
is that the inclusion of time is necessary to have a meaningful interpretation of
reciprocity. After all, reciprocation is a notion of giving back. But in an important
class of gift-giving, namely charity, gift-giving occurs anonymously and therefore
necessarily without any countergift.
Quite obviously does the absence of countergifts not preclude gift-giving. The
difficulty therefore lies in the anonymous aspect of charity. Can approval be obtained in such a setting? The answer, perhaps surprisingly, is sometimes yes.
Status is often awarded on the basis of events that do not directly affect the
person who awards status (Coleman [1990, 130]). The receiver may very well be
anonymous, but the giver can make his act known to other persons than the receiver. This is most clearly described by Schwartz [1967, 2] who states that “it
is common knowledge that men present themselves publicly by the conspicuous
presentation of gifts. Generous contributions to charity have always been a source
of prestige in the United States.” This is acknowledged by fund raisers by providing people with “I gave” stickers to be affixed to the front door (Schwartz [1967])
or coffee mugs, lapel pins, and bumper stickers and so on (Andreoni and Petrie
[2000]). In this way, donors can signal their contributions.
Clearly, if approval is indeed the motivation of charity then behavior should
depend heavily on whether or not the possibility exists to make the act of giving publicly known. In a clever designed experiment Andreoni and Petrie [2000]
have tested for this. They found that, if faced with the opportunity, virtually all
subjects that donated choose to be publicly known as having done so. Moreover,
average contributions in last rounds are more than twice as high as in anonymous
settings. Satow [1975] has found a similar effect in an experiment where subjects
could voluntarily donate part of their earnings to a research fund. Not only did
they donate significantly more to the research fund when they were observed by
the experimentator (public condition) than when they were unobserved (private
Chapter 3. The Demand for Social Approval as a Motivation to Give
72
condition), the amount donated was also significantly related to their need for
approval, as measured by a ’social desirability scale’. Subjects with a high need
for approval donated significantly more than those with a low need for social
approval. Furthermore, there was also a significant interaction effect: under the
public condition the effect of a high need for approval was greater than under the
private conditions. The main results are reproduced in the table below. Unfortunately, I am not aware of any other experimental study that relates gift-giving to
the need for approval, so the evidence remains circumstantial.
TABLE 3.1
Mean Percentage of Earnings Donated to the Research Fund
Need for approval
Private condition
Public condition
Low need for approval
4.45
17.27
High need for approval
3.11
38.71
Source: Satow [1975].
As a last piece of evidence, I report an empirical study by Harbaugh [1998].
He found that when charities report names of contributors in categories rather
than in exact amounts, a large proportion of donations is at the boundaries of
those categories. Social approval as a motivation predicts this: it is impossible
for the public to discriminate between different donations within each category.
The donor can therefore as well contribute the lowest amount within a certain
category.10
3.2.4 Gifts and markets
Thus far the focus has been on the properties of gift-giving. In this subsection
these findings are related to another exchange mechanism: the market.
Recall once again that gifts in kind are usually taken to be a social waste because the recipient could have achieved a higher utility level from an equally costly
cash-gift. Problematic in this respect is that in the environment of gift-giving, individual rationality sometimes prescribes to give in kind rather than in cash (see
10 Though
a reasonable alternative explanation is that framing effects play parts. Note also that the significance of this evidence is undecisive as there may be many more institutions that do not publish names.
73
Social approval, status, and gift-Giving
proposition 2). The market mechanism does not suffer from this tendency. This
has warranted researchers to conclude that, compared to the market, gift-giving
creates a deadweight loss. Subsequent calculations accredited to Christmas gifts
a deadweight loss in the range of 10-30% of their value (Waldfogel [1993]).
In the research done by Waldfogel [1993] and in the follow-up studies (Solnick
and Hemenway [1996] and Waldfogel [1996]) it is tacitly assumed that the giver is
indifferent between giving in cash and in kind. However, as found in questionnaires
(see e.g. Webley and Wilson [1989]) and in line with Proposition 2, givers are not
at all indifferent but clearly prefer to give in kind. This has consequences for
the measurement of the deadweight loss. True, turning to the market prevents
inadequacy. But insofar as the market is an anonymous institution, it does not
realize this valuable psychological sentiment called approval. It is for that reason
not obvious that gift-giving creates a deadweight loss. The trade-off to be made is
that between adequacy and approval. In neglecting the latter part of this trade-off
one tends to overestimate the size of the deadweight loss, if there is any.
Of equal interest is how the market institution interacts with gift-giving. There
are instructive examples that show how the opening of a market leads to a
crowding-out of gift-giving (Hirsch [1976], Yellen [1990]). There are hints that
this crowding-out is detrimental to social welfare. The keypoint is that the links
between the market institution and gift-giving are shot through with externalities
(see in particular Kranton [1996a] and Dasgupta [2000], and chapter 4). As Hirsch
[1976, 78] puts it, “social relationships do not, by their nature, have the character
of private economic goods.” Indeed, when the !Kung tribe gained access to markets which were superior in the supply of goods, they abolished gift-giving and
at the same time retreated from their social life, eroding the cohesion of society
(Yellen [1990]).
Yet, there is also another side to the coin. Where the market is an efficient
institution, gift-giving can be a hindrance. Kranich [1994] points out that when
gifts are permitted, an efficient equilibrium need not be reached. For instance,
gift-giving acquires the character of a public good in case agents have preferences
over the economy’s entire allocation of income rather than just their own income.
As another example, recall that due to moral hazard the market does not provide full insurance. Nonmarket insurers (such as a family) may therefore want
to supplement the market. Arnott and Stiglitz [1991] then show that when these
nonmarket insurers have no better information than market insurers, reciprocal
Chapter 3. The Demand for Social Approval as a Motivation to Give
74
assistance not only crowds out market insurance but is also harmful and therefore
dysfunctional.
On the more positive side can gift-giving be a useful complement to the market
institution in case the latter fails. Consider the supply of a public good. The
market creates no incentives to contribute at all because individually it does not
pay off in terms of consumption. Social approval at least gives some incentives,
although there is nothing that guarantees that gift-giving for the sake of obtaining
approval continues up to the point where the supply is optimal. Gift-giving can
complement the market to a certain extent, though additional incentives are likely
to be needed in order to reach the optimal amount of gifts.
In providing these additional incentives, it is probably not unfair to say that the
effects on social approval are usually neglected when it comes to institutions being
recipients. The right way to stimulate giving then seems obvious: compensate
donors directly for the efforts. This is exactly the measure that was expected to
alleviate the shortage of blood donations. However, quite unexpectedly to many,
the actual change in the supply of blood turned out to be negative (see Titmuss
[1970]). Other studies that show this pattern emerges in different settings as well
include Gneezy and Rustichini [2000] and Frey [1997a], [1997b] (see also chapter
7). Yet, this kind of behavior is only surprising insofar as the object of human
behavior is material gain. A higher compensation makes you richer and it makes
giving relatively cheaper. For that reason one should indeed, if anything, increase
one’s gift. On the other hand is there the effect on social approval. Rewarding
gift-giving reduces sacrifice, and as argued before, there is evidence that sacrifice
is positively correlated with social approval. Hence, rewarding gift-giving deprives
the agent from the opportunity to realize social approval. This latter effect reduces
the incentives to give and makes the end outcome ambiguous.
Consider therefore the case where the receiver does not make a countergift but
gives a monetary compensation instead. Thus, set l2v = 0 and denote by wv the
compensation the institution offers for each unit of voluntary labor. The forgone
wage of giving is now w̃i ≡ wim −wv > 0. I say that sacrifice is relatively important
if the approval rate is sensitive to the sacrifice made: i.e. ∂β i /∂ w̃i is high.
Proposition 4 (Compensation): With a preference for social approval, the optimal gift size of player 1 is decreasing in compensation (dl1v /dw v < 0) if the
Social approval, status, and gift-Giving
75
absolute elasticity of u0s is less than unity and if sacrifice is relatively important
(∂β 1 /∂ w̃1 high).
Corollary 1 If compensation crowds out gift-giving then welfare is reduced:
∂u1 /∂wv < 0 if dl1v /dwv < 0.
Proof. With l2v = 0, player 1’s utility function becomes (ignoring fixed parameters): u1 = u1 (l1v (wv ), w v ). Differentiating at the optimum with respect to w v
2u
dlv
0 ∂ 2 x1
00 ∂x1 ∂x1
0 ∂ 2 s1
1
gives: sign of dw1v is equal to the sign of ∂l∂v ∂w
v = ux ∂lv ∂w v + ux ∂w v ∂lv + us ∂lv ∂wv +
1
∂s1 ∂s1
u00s ∂w
v ∂lv . The first two effects are positive. Since
1
1
∂ 2 s1
∂l1v ∂w v
1
1
= − ∂∂βw̃11 the last two
2
u1
effects can be combined into − ∂∂βw̃11 (u0s + β 1 l1 u00s ). Clearly, ∂l∂v ∂w
v < 0 requires that
1
0
00
0
us + β 1 l1 us > 0 (which is true for an absolute elasticity of us smaller than one),
and that ∂∂βw̃11 > ξ ≡ [u0x − w̃1 l1 u00x ] / [u0s + β 1 l1 u00s ] . This proves Proposition 4. To
prove Corollary 1, note that du1 (l1v (wv ), wv )/dwv
= (∂u1 /∂l1v )·(dl1v /dw v )+(∂u1 /∂wv ) = ∂u1 /∂wv |l1v∗ . Hence, at the optimum, utility
decreases in compensation if ∂u1 /∂wv = l1v u0x − (∂β 1 /∂ w̃1 )l1v u0s < 0 ⇔ ∂β 1 /∂ w̃1 >
u0x /u0s . With the above definition of ξ we have ∂∂βw̃11 > ξ > u0x /u0s so that if beta is
high enough such that compensation reduces gift-giving, then compensation also
reduces welfare.
The intuition is again straightforward. If sacrifice is relatively important then
every unit of voluntary labor is valued a lot less when sacrifice is a little bit
lower. This reduces the incentives to give if compensation is offered. There is also
a positive effect on social approval since a lower β reduces s and this increases
marginal utility of giving. But this latter effect is always dominated by the former
effect as long as the absolute elasticity of u0s is smaller than one. Furthermore,
by Corollary 1 the deprivation of social approval is in this case enough to reduce
utility.
A result similar to proposition 4 is obtained by Holländer [1990] who argues
that the opening of a market reduces incentives to give to a public good. If the
market (or government) provides some of the public good, then the marginal
benefits of the public good declines and under his specification this means that
less approval is obtained for each contribution.
It should be emphasized that crowding-out is not a general phenomenon, but
Frey and Jegen [2002] conclude that there exists compelling empirical evidence
that it does occur on some occasions. To appreciate the significance of this to
Chapter 3. The Demand for Social Approval as a Motivation to Give
76
its fullest extent, note that gifts in the form of bequests can affect the longrun growth of an economy (see Galor and Zeira [1993]). The insight gained is
therefore of importance, so much the more Frey [1997b] argues that in addition
there are spill-over effects.11 For example, spill-over effects to other activities or
spill-overs over time may occur. Thus, the social approval that can be obtained for
giving this particular good can influence the approval for giving some other good.
Or, not unreasonably, social approval is time dependent.12 Giving an inadequate
gift reduces social approval now, but persistently giving inadequate gifts reduces
social approval even more in the future. Albeit these are interesting extensions
of the model, it would be too speculative to elaborate upon them at the present
state. More should be known about the exact determinants and shape of the
social approval function, as well as its dynamics. Nonetheless, the conclusion
can be drawn that one should be very careful with the implementation of market
incentives in non-market institutions. This can result in unintended consequences
that are harmful to social welfare.
3.3 Related literature
The idea of social approval seems intuitively plausible. It is therefore no surprise
that social approval takes a prominent place in other social sciences. This makes
it all the more remarkable that the economics literature is mostly silent on this.
The works that are most closely related are that of Holländer [1990], Frey [1997a],
[1997b] and Andreoni [1990]. But there are some notable differences. Holländer
relies on the seemingly unintuitive assumption that sacrifice is negatively correlated with approval. Compensation would then increase gift-giving, contrary
to many empirical findings. Frey [1997a], [1997b] and Andreoni [1990] use other
emotive concepts that are very close to approval, namely intrinsic motivation and
warm glow respectively. Both concepts attribute the act of gift-giving to the fact
that people derive utility from the act of giving per se. To a certain extent these
concepts are compatible with the current framework.13 The advantage of the cur11 The
approach of Frey [1997a], [1997b] to explain crowding-out is taken up in Section 3.3.
of the evidence reported in Frey [1997b] as well as in Gneezy and Rustichini [2000] indicates persistence over time.
13 One difference is that the concepts of intrinsic motivation and warm glow feeling also work in a strictly
anonymous setting whereas social approval does not. But see the discussion in Section 2.3 where it is argued
that even charity is often not truly anonymous.
12 Some
77
Related literature
rent framework is that an attempt is made to further uncover the exact channel
through which the intrinsic motivation or a warm glow feeling is obtained. This
leads to a better insight in what the properties of gift-giving and the gift itself
will be. For example, if social approval is indeed the real source of the warm glow,
then the assumption of Andreoni [1990] that the warm glow is independent of the
subsidy is doubtful since a subsidy reduces sacrifice. This doubt is strengthened
by the fact that this assumption rules out a negative relation between subsidy
and gift-giving, contrary to the empirical findings. Neither do the theories of Andreoni [1990] and Frey [1997a], [1997b] in their current state have anything to say
about the optimality of the adequacy of the gifts even though this property is of
critical importance in making welfare judgements. Central in the theory of Frey
[1997a], [1997b] is that intrinsic motivation is crowded out by extrinsic motivation. Although this may be costly, the extrinsic motivation perhaps results in a
higher level of adequacy.
There is by now a fast-growing literature that tries to explain the properties of
gift-giving through other causes than social approval, see in particular chapter 2.
These include altruism, signaling, and fairness. Alike the current approach, each
of these alternative theories has its own deficiencies. For example, if altruism is
the motivation to give as suggested by Becker [1974], then the higher the utility of
the recipient the better, so that inadequate gifts cannot be explained. Neither can
crowding-out be explained, since a compensation would make gifts less effective
so more would be given to obtain the desired distribution. On the other hand,
with social approval compensation creates costs to the giver since he is deprived
from some approval and this cannot be costlessly compensated by giving more.
In case information is asymmetric, gifts can also act as a signal. Exemplary
in this respect is the work by Camerer [1988]. In his model, honest players can
make gifts to signal themselves as being trustworthy. Cheaters do not find it
profitable to make gifts. As a corollary, this explains the inadequacy of gifts,
because with adequate gifts cheaters would participate in the gift-giving game
anticipating that they will collect valuable gifts. However, a compensation would
increase gift-giving since it would make current gifts less effective as signals. More
should be given to reveal one’s type.
The fairness approach has probably gotten most attention in the literature (see
especially Fehr and Schmidt [1999], and Bolton and Ockenfels [2000]). It successfully accounts for a broad range of experimental games by assuming that people
Chapter 3. The Demand for Social Approval as a Motivation to Give
78
not only care about their own monetary payoffs, but also about the distribution
of payoffs and the intentions that other players have. The theory is especially
productive in explaining reciprocity: just think of the saying “an eye for an eye,
a tooth for a tooth.” Inadequacy seems not to be easily explicable: if one gives
to make the distribution more equal, then the most effective way to do this is
to give as adequately as possible. People would therefore prefer to give adequate
gifts and only give inadequately if no better alternative is at someone’s disposal,
e.g. if the only choice they face is to give inadequately or not to give at all. For
similar reasons should a compensation increase gift-giving: more should be given
to achieve the desired equality in the income distribution.
As it stands, no single theory can explain everything. The current approach
should therefore be considered as being complementary to the existing literature,
not as an alternative theory. Future research will aim at synthesizing the different theories, for example by arguing that people are intrinsically motivated by
principles of fairness and at the same time also want to be seen as such, that is,
to be approved by others for being fair.
3.4 Concluding remarks
The purpose of this chapter is to argue that at least part of gift-giving can be
sensibly understood as the result of a desire for social approval. At the cost of
a loss in predictive power, the analysis is kept rather general. The motivation
for that is simply because little is known about the exact shape of the approval
function. However, under weak but nevertheless (at least in my view) reasonable
assumptions I have been able to derive some qualitative properties that I think
are not sensitive to the specifications of the model.
The foregoing shows that social approval may indeed play a role in people’s
behavior. This is not to say that social approval is the unique motivation next to
one’s own payoff or that social approval plays a role in every kind of exchange.
Indeed, it is very probable not the unique motivation. The gift-giving behavior
found in truly anonymous games cannot be attributed to a taste for social approval. But insofar as it does explain part of the behavior, its implications are
far reaching. Wherever social approval is important the superiority of the market
institution to gift-giving must be reconsidered, despite the fact that gift-giving
elicits inadequate gift exchanges. And where the market is considered to be su-
79
Concluding remarks
perior but does not arise from spontaneous order, the incentive schemes based on
price incentives as designed by typical economic arguments should be rethought
so as to take into account the effects they have on social approval. A better solution might be to make the gifts more visible to outsiders, rather than to give
monetary compensations.
Prior to that, more research is needed to assess when social approval is a main
driving spirit, what the exact determinants are, and furthermore to substantiate
the role of inadequacy of gifts as a strategic variable. Perhaps a promising direction to go in is to design experiments that not only reveal the identity or actions
of players, but in addition communicate more precise information on the approval
of players. Because the propositions are stated in terms of how social approval is
affected by different actions, information on the approval allows the propositions
to be more specifically tested. More concretely, one possibility is to include an
index of how each player approves the gift. This allows players to reveal their appreciation of the gift. It is then possible to see how the other player responds to
changes in that index. If this index is designed carefully, the propositions in this
chapter lend themselves to being tested. Proposition 2 can be tested by making
the adequacy of the donation in a public goods games a choice variable. This is
in the spirit of Goeree, Holt, and Laury [2002]. They test the effects of changing
the external rate of the contribution. The external rate is the rate at which other
players profit from your gift, the adequacy level in other words. In their experiments the external rate is a given variable for the subjects. Finally, Proposition
4 can be tested by examining whether the existence of crowding out is related to
the approval obtained in a way corresponding to the conditions stated.
4
On the Viability of Gift-exchange in a Market
Environment
In the harsh Kalahari Desert in Africa, a group of people known as the !Kung
lived for a long time by the principles of reciprocity and sharing. These principles
provided enough security not to live on the brink of starvation, despite the apparent
difficult circumstances. Then, the access to the market caused an influx of money
and goods. This was partly responsible for the eroding traditional values and social
cohesion. To which extent can we expect that, over time, access to the market
crowds out reciprocity?
4.1 Introduction
Reciprocal exchange in its pure form can be observed in special places where the
market is not strong enough to break personal connections. On the other hand,
there are fascinating stories by anthropologists showing how reciprocal exchange
arrangements vanish when tribes encounter markets. This chapter studies the
relationship between market exchange and reciprocal exchange.
In an interesting paper Kranton [1996a] shows that in order to become beneficial, markets need enough participants to reduce search costs. Therefore, reciprocal exchange may survive if initially the proportion of the people that engage
0 This
chapter is coauthored by Theo van de Klundert. We are indebted to Jan Boone, Lans Bovenberg,
Patrick François, Richard Nahuis, and Sjak Smulders for useful comments on an earlier version of the paper.
Chapter 4. On the Viability of Gift-exchange in a Market Environment
82
in market exchange is not too large. There are other determinants which may tip
over the balance in favor of market exchange, such as a lack of trust. In reciprocal
exchange people have to trust each other, because production and consumption
are separated over time. Moreover, reciprocal exchange may involve a less elaborate division of labor.
Kranton [1996a] borrows evidence documented by Yellen [1990] that describes
how the !Kung tribe abandoned reciprocal exchange once they encountered the
market economy of Botswana. But reciprocal exchange is not limited to tribal
communities nor to some corners of the economy but is an element of importance in developed market economies as well. Reciprocity or gift exchange is an
essential aspect of culture. There are social and moral dimensions to economics
so to say (see for instance Etzioni [1988]). Even Adam Smith already knew that
”moral sentiments” are important, something also recognized by Kenneth Arrow
: ”ethical behavior can be regarded as a socially desirable institution which facilitates the achievement of economic efficiency in a broad sense” (Arrow [1972,
354]). People have a sense of belonging to society at large, and care for social
interactions. This induces cooperative behavior in different guises. Self-interest
seeking behavior is a sine qua non for coordination in the economy, but there are
limits to opportunistic behavior. People want to be respected by others. To a certain extent respect follows from success in the accumulation of wealth, but there
are limits to respectfulness in this sense. Mutual aid and sympathy are important
values of their own. For this reason producers may take pride in the quality of
the product they deliver, workers may be motivated to do a good job, and people
in general may take account of each other’s interests in different situations. Experiments in economic settings indeed confirm that people may behave different
from what standard neoclassical economics predicts. Take for instance, Gächter
and Fehr [1999] who find that social approval and social familiarity generate a
significant rise in cooperative behavior (see also chapter 3).
Casual observation learns that people are concerned about a loss of commitment in recent times. This is often associated with commodification, meaning that
markets are expanding into almost every territory of human life. In our interpretation this means that the market system crowds out reciprocal exchange, bringing
about a lack of close personal relationships. The questions to be answered are
then the following. What are the main factors causing such a crowding-out? Is
reciprocal exchange in the sense of moral behavior completely wiped out or are
83
Introduction
there conditions such that both regimes can coexist in a long-run equilibrium?
To answer these questions we apply the analytical framework based on Kranton
[1996a] and Diamond [1982] in a modified way, and with a different interpretation.
In the model used, people can make a deliberate choice between two regimes: a
reciprocal relationship or market exchange. In reciprocal exchange agents value
social behavior as they have a sense of belonging to society at large. Applying the
terminology introduced by Khalil [1997], it can be argued that agents produce
goods from which they derive substantive utility, but sticking to the moral codes
of the group provides also satisfaction in the form of symbolic utility. Initially
we assume that the terms of trade between substantive and symbolic utility are
deteriorating over time. Extending the model later on, it will be argued that the
terms of trade typically depend in a non-linear way on the fraction of agents
engaged in reciprocal exchange.
In the regime of market exchange agents behave as prescribed in neoclassical theory. If the fraction of agents engaged in pure market activities rises, the
market operates more efficiently. People are then less hampered by tradition and
can seize every opportunity for making a profit. The division of labor can be exploited more fully. The set of goods produced generally differs from that in case
of reciprocal exchange. This has an impact on substantive utility. The idea that
market efficiency is related to the number of people in the market is characteristic
for search models. We generalize this idea by assuming that more encounters between people induce more production by leveling organizational and institutional
restrictions. However, both ideas can be modelled in the same manner.
The model applied here differs in a number of aspects from that in Kranton
[1996a]. First, in modelling the search externality we follow the original set-up
of Diamond [1982], [1984] more closely. Second, to simplify further we assume
that goods are produced at constant cost instead of introducing a distribution
from which agents draw randomly. Finally, it should be observed that we not only
obtain corner solutions as in Kranton. Assuming that the terms of trade between
regimes depend on the fraction of agents engaged in the market system we find
interior solutions with agents operating under different regimes.
The rest of this Chapter is organized as follows. In section 4.2.1 we first specify
what is meant by sympathy and substantive utility. This sets the stage for modelling reciprocal exchange in section 4.2.2. Market exchange is discussed in section
4.2.3. In section 4.3 we consider equilibrium solutions under different assumptions
Chapter 4. On the Viability of Gift-exchange in a Market Environment
84
with respect to the shape of symbolic utility. Complete commodification obtains
in section 4.3.1. Interior solutions with partial commodification are considered
in section 4.3.2. The role of discounting is scrutinized in section 4.3.3. Welfare
considerations are taken up in section 4.4. The chapter closes with concluding
remarks in section 4.5. Proofs of propositions are deferred to the appendix.
4.2 Exchange mechanisms
In this section we describe the formal model. The general setup is one in which
each agents starts in a situation in which he is involved in a personal relationship
with one other agent. A key feature of such a reciprocal exchange relationship is
the element of trust. Production and consumption are typically separated over
time. Contrary to the market where money serves as a medium of exchange, no
such security exists in a reciprocal relationship. It is therefore possible that agents
who did produce last period see their relationship end without having the possibility of consuming in return. Although this favors the existence of markets,
market exchange has disadvantages of its own. The market is typically characterized by anonymous agents, without relation-specific commitments, and search
costs have to be made in order to find a trading partner. On the other hand is it
generally acknowledged that the market is capable of supplying a larger array of
goods.
Before turning to a detailed exposition we present an outline in figure 4.1. This
facilitates the reading of the subsequent sections. The arrows denote possible
flows. Agents start in a reciprocal exchange relationship. They can end their
relationship and enter the market as unemployed. The distribution of agents in
each period is given by the fractions r, u, and m that correspond to the different
states as in figure 4.1. The market is characterized by a search process that
describes how agents become employed with a good, sell that good and become
unemployed again. This search process is determined by the technical parameters
a and b to be explained later on.
We start with an exposition of reciprocal exchange and postpone a treatment
of the market part to the next section. First however, we discuss the agents’
preferences in more detail.
Exchange mechanisms
85
Market exchange
Reciprocal
exchange
(r)
a
unemployed
(u)
employed on
market (m)
b(m)
FIGURE 4.1.
4.2.1 Substantive and symbolic utility
Following Khalil [1997] we distinguish between substantive utility (ordinary tastes
for material goods) and symbolic utility (tastes for selfhood and alike). Consumers
derive substantive utility from a basket of goods, which may differ across regimes.
Substantive utility for each representative consumer is denoted by xi , where the
index i = r, m denotes reciprocal or market exchange. Symbolic utility can be
taken into account by introducing extended preferences (see for instance Becker
[1996]). Here we take a short-cut by introducing a mark-up (θi ) on substantive
utility. Extended utility is then defined as:
yi = θi xi , θ i ≥ 1, i = r, m.
(4.1)
It is assumed that under market exchange extended utility coincides with substantive utility (θm = 1). Reciprocal exchange conveys symbolic utility as people
value social interaction as such (θ r > 1). The above set of equations can be
combined to:
yr = θym ,
(4.2)
where θ ≡ (θr xr )/xm . As substantive utility is fixed xr and xm can be treated as
constants. It seems plausible to assume that θr , and therefore θ, depends on the
Chapter 4. On the Viability of Gift-exchange in a Market Environment
86
fraction of people engaged in market exchange, m. If the number of people in gift
exchange declines it will be harder to uphold a sense of community spirit. As the
market regimes expands, individualism spreads and solidarity may become less
attractive. These considerations lead to a negative relation between the valuation
ratio θ and the fraction of people in the market regime. However, according to
Adam Smith it can be maintained that the need for mutual sympathy and respect is deep-rooted and cannot be suppressed entirely. A similar view based on
biological principles is expressed in Kropotkin [1904]. It is not unreasonable to
assume that the ramification of this becomes more distressing as the market gets
to dominate exchange relations. This could imply a positive relation between θ
and the market size after some threshold level of market participation has been
passed. Thus we have that θ = θ(m) is first decreasing and after some critical
point, say m0 , increasing, because the market becomes ”too large”.
More specifically, the U-shaped θ(m)-curve can be seen as the result of several
opposite forces. For instance, as more people leave the regime of reciprocity the
cost (in disutility terms) of changing beliefs for the remaining people decline.
This idea builds on the literature on cognitive dissonance in psychology. Following
Festinger [1957] it can be stated that changing beliefs induces a negative arousal.
Moreover, the resistance to change crucially depends on the difficulty of finding
people who support the new belief. Therefore, the more people are already in
the market regime the easier it becomes for people to switch regimes. It is in
particular hard for the first few individuals that are to switch. This is shown by
the dashed downward sloping RC-curve (resistance to change) in figure 4.2 .
Furthermore, as the market system expands the social deficit becomes more
important. As argued in Bowles [1998] markets are characterized by impersonality
and ephemerality of contact. But people also want to socialize. There is a need for
mutual sympathy and recognition. If reciprocal exchange is relatively large market
participants may have the feeling that the social deficit can be easily repaired.
The more people are available as potential candidates to socialize with, the easier
it is to change back to reciprocal exchange. Therefore, the opportunity costs (in
disutility terms) of switching to the market regime increase as the market system
becomes dominant. The social deficit is felt more heavily for a larger market share.
This gives rise to the dashed upward sloping SD-curve (social deficit) in figure
4.2 . Under appropriate conditions the summation of both curves may lead to an
U-shaped function θ(m) as illustrated by the bold curve in figure 4.2. To analyze
Exchange mechanisms
87
different possibilities, the equilibrium solution discussed in section 4.3.1 will be
based on a monotonic decreasing θ-curve. This case eventuates in the corners as
the only solutions. The U-shaped θ-curve will be introduced in section 4.3.2. As
it turns out, this opens the possibility of interior solutions.
2(m)
RC
SD
m
FIGURE 4.2.
It should be noted that there is a close parallel with the descriptive approach
of Fukuyama [1957]. In his latest book Fukuyama describes ”the great disruption” of the social system starting somewhere in the sixties under influence of the
upcoming information technology. The change of a traditional industrial society
towards an economy dominated by the service sector leads to a certain disorientation and as a consequence of this to a decay of moral values. But according to
the author things have changed lately. The reason is that human nature is geared
towards cooperation and reciprocity. Fukuyama bases his view on the biological
approach going back to Kropotkin [1904]. The reconstruction of the social order
is reflected in declining criminality statistics and a more positive evaluation of
social relations in systematic surveys. Such a change may lead to a new equilibrium with a certain amount of gift exchange in the current market economy. In
our model it is the upward sloping branch of the θ(m) curve which reflects the
Chapter 4. On the Viability of Gift-exchange in a Market Environment
88
views of Fukuyama and others on the viability of the moral system in modern
times.
4.2.2 Reciprocal exchange
The timing in a reciprocal exchange relationship is as follows. Each two periods a
complete reciprocal gift exchange can be accomplished. In the first period, one of
the agents produces first and the other consumes the good. In the second period
they switch roles. Whenever they consume, they derive utility yr , and whenever
they produce they bear a cost in disutility terms of c1 . The discount factor for
the next period equals δ = e−ρ , where ρ is the subjective discount rate2 . Agents
are infinitely lived.
Let Vrp and Vrc denote the lifetime discounted utility of the agent that is involved in reciprocal exchange and starts as producer and consumer respectively.
The agent starting as a producer incurs a cost (c) and expects to be in the position
of a consumer next period:
Vrp = −c + δVrc .
(4.3)
Similarly, the lifetime discounted utility of the agent that starts as consumer, Vrc ,
is given by:
Vrc = yr + δVrp .
(4.4)
It is further assumed that each agent ends up being first consumer or producer
with equal probability. Expected discounted lifetime utility of staying in a reciprocal exchange relationship is then given by Vr = 12 (Vrp + Vrc ) or, after substitution
and rearranging terms:
yr − c
Vr =
.
(4.5)
2(1 − δ)
The element of trust is introduced by the possibility of ending the relationship.
An agent can decide to consume first, but not to produce in return. He then
consumes and enters the market. Whenever he does, the other agent will of course
take notice of being cheated and as a punishment he will end the relationship3 . As
1 For
simplicity we assume that the complete production of an individual is exchanged. A more realistic
extension would be that individuals exchange only part of their production.
2 A natural restriction on the discount factor is that δ ∈ (0, 1): future revenues are valued positively but less
than current revenues.
3 This tit-for-tat is only one of many possible strategies. The strategy is common in the microeconomic
literature. The classic defence is given in Axelrod [1984]. More interestingly, the strategy is defended by Aristotle
on principles of justice: ”Now if proportionate equality between the products be first established, (...) then
89
Exchange mechanisms
a consequence, both agents will have to enter the market. Obviously, the agent
will deliberate upon producing in order to sustain the relationship or to cheat
upon his partner. Whether the agents decide to cheat and enter the market or
stay in the reciprocal exchange relationship is dependent on the derived utility
of being in the market regime, Vu (to be specified later on), and of the derived
utility in the reciprocal exchange relationship, Vr . The following definition shows
a Nash-equilibrium constraint under which both agents will decide not to cheat
(a derivation is given in the appendix).
Definition 1 (enforceability) A reciprocal exchange relationship is enforceable if
−c + δVr (·) ≥ Vu (·).
The constraint is more likely to be satisfied when the discount factor is high (low
subjective discount rate) or when the market size is small. The higher the future
is valued, the less beneficial it is to cheat and so reciprocal exchange is more easily enforceable. Alternatively, we can interpret the subjective discount factor as a
measure of trust. If the discount factor is low, then the faith in getting consumption in return is poor. The lower the discount factor, the harder it is to uphold
the relationship. The market, on the other hand, is characterized by increasing
returns to scale. The larger the market, the more easy it is to find a partner to
trade with. This implies that the value of trading on the market is positively dependent on the market size4 . The enforceability constraint is therefore harder to
satisfy at larger market sizes. Detailed comparative statics are provided in section
3.3. We now turn our attention to the determination of the value of entering the
market.
4.2.3 Market exchange
Agents that have decided to enter the market do not have a fixed trading partner.
They enter the market ”unemployed”, that is with no goods, and have to search
for production possibilities. This will be represented by a Poisson process with
arrival rate a. They can either accept or not accept the production opportunity.
After they have found and accepted one they bear the same disutility costs c as
in reciprocal exchange. Being employed they still have to search someone to trade
reciprocation take place (...) but if this is not done, the bargain is not equal, and intercourse does not continue.”
(Aristotle [1994, 283]).
4 From (4.9) introduced later in the text, it is easily proved that V is indeed (weakly) increasing in m.
u
Chapter 4. On the Viability of Gift-exchange in a Market Environment
90
with. They find someone with probability b(m). Having found a trade partner they
exchange and derive utility ym .5 The market is characterized by increasing returns
to scale in the form of an externality. As the fraction of the population on the
0
market, m, gets larger, average search time decreases or equivalently b (m) > 0.
Recall that r, u, m are the fractions of the population in reciprocal exchange,
unemployed on the market, and employed on the market respectively (see also
figure 4.1). If we normalize total population to unity then u = 1 − r − m. The
flow dynamics in the market can then be described by the differential equation:
ṁ = a(1 − r − m) − b(m) · m.
(4.6)
Here, ṁ ≡ dm/dt. The fraction of people on the market increases with the number
of agents finding a production possibility and decreases with the number of agents
accomplishing their exchange.
In the remainder of the chapter the focus is on steady state solutions. Steady
states are marked by a constant distribution of agents over states, hence we
have a constant rate of employment: ṁ = 0. Based on this assumption we can
derive the value equations of the agents on the market. Let Vu and Vm denote the
discounted lifetime utility of being unemployed and employed respectively. Under
the assumption of a steady state, dVu /dt = dVm /dt = 0 and the value equations
are given by:
ρVu = a(Vm − Vu − c),
(4.7)
ρVm = b(m)(ym + Vu − Vm ).
(4.8)
This set of equations can be rewritten in terms of Vu . Since unemployed agents on
the market always have the possibility of not accepting a production possibility,
due to high costs, Vu is always nonnegative6 :
½ ·
¸ ¾
a b(m)(ym − c) − ρc
,0 .
(4.9)
Vu = max
ρ
ρ + b(m) + a
definition 1 is now completely determined by equations (4.2)-(4.9). Together with
the steady state condition that ṁ = 0 this describes the long-run equilibrium.
5 The market is typically characterized by the use of money. Diamond [1984] explicitly takes money into
account by distinguishing between unemployed, buyers, and sellers. Here we assume that buying and selling
simultaneously take place. None of the results are sensitive to this assumption.
6 We exlude negative values of V by assuming that agents can choose for complete idleness with V = 0.
u
u
91
Equilibrium
4.3 Equilibrium
Based on the steady state assumption, this section explores the consequences of
different assumptions on the shape of the valuation ratio θ(m). We stay close to
the disquisition in the introduction and section 4.2.1. Thus in section 4.3.1 we
examine the steady state solutions when the valuation ratio θ is monotonically
decreasing in the market size and in section 4.3.2 we allow for an increasing part
of the θ-curve for large market sizes. Finally, comparative statics are provided in
section 4.4. Throughout we use the notion of a short-run equilibrium whenever
ṁ = 0, and of a long-run equilibrium when ṁ = 0 and, in addition, definition 1
is satisfied.
4.3.1 Complete commodification
Where markets replace social relations we speak of commodification. Complete or
full commodification indicates a situation in which markets expand to an extent
where all social relations are abolished. Under incomplete or partial commodification social relations are still embedded in the community, but are partly driven
out by the existence of markets. In this section we consider the case where θ
is monotonically decreasing in the market size (for reasons given in the introduction and section 4.2.1). As it turns out, this is a situation where, if any, full
commodification results.
Figure 4.3 depicts this case7 . The Er and Eu curves show respectively the LHS
and the RHS of the enforceability constraint in definition 1. Therefore, whenever
the Er lies above the Eu curve reciprocal exchange is enforceable, and for market
sizes where Eu lies above Er agents maximize expected utility by entering the
market. The value of being unemployed increases in the market size and is therefore upward sloping. The value of reciprocal exchange depends on the valuation
ratio θ(m) and is therefore downward sloping. The curves are drawn for the range
[0, m̃]. These are the possible short-run equilibrium market sizes. The upper limit
is given by m̃; the maximum possible market size for which no agents are involved
in reciprocal exchange, i.e. r = 0.8 The critical value of the market size at which
7 The following set of parameters is used to obtain the figure: {a, b, c, g, h, y , δ} = {.8, .6, 1, −.1, .3, 8, .8}
m
where we assumed that the probability of finding a trading partner on the market is linear in m: b(m) = b · m
and the subjective valution is given by θ(m) = h + gm. This set of parameters assures that both exchange
mechanisms are enforceable for some market sizes for a value of θ in the range: θ < θ < θ. (see section 3.3).
8 Thus from equation 4.6 and the definition of short-run equilibrium m̃ is such that a(1 − m̃) = b(m̃) · m̃.
Chapter 4. On the Viability of Gift-exchange in a Market Environment
92
the enforceability constraint holds with equality is given by m1 . At all market
sizes m ≤ m1 reciprocal exchange is enforceable. Whenever the market size for
some reason exceeds the critical size m1 the market size is too large for reciprocal
exchange to be enforceable and eventually all agents will enter the market. This
cannot be a long-run equilibrium. The only possible long-run equilibria are when
economy ends up in a corner solution9 . As appears from equation (4.6) the longrun equilibrium in case people opt for the market regime equals m̃. Thus we see
that for some market sizes reciprocal exchange is enforceable and that for larger
market sizes the economy will converge to a market exchange economy. It is clear
from the picture that if the Eu curve is everywhere above the Er curve then reciprocal is never enforceable no matter what the market size is. Similarly, if the Eu
curve is everywhere below the Er curve then reciprocal is always enforceable.10
6
4
Eu
2
Er
0
-2
0
0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9
m1
1
m
~
m
FIGURE 4.3.
9 Strictly speaking, it is possible that the economy sticks to its initial market size (possibly at a positive
level) or converges to a market size of m̃.With slight abuse of notation, we speak of corner solutions even if the
initial market size is positive.
10 However, V = 0 at m = 0 and if E is everywhere below E then E must be negative. But V can still
u
r
u
r
r
be positive and the case is therefore meaningful.
Equilibrium
93
4.3.2 Partial commodification
In the previous section, the valuation of reciprocal exchange was assumed to be
monotonically decreasing in the market size. However, as mentioned earlier (notably in section 4.2.1) it is more likely that the valuation depends on the market
size in a slightly more sophisticated way. Despite the decreasing disutility costs
of changing beliefs, the feeling of a social deficit that cannot be repaired becomes
distressingly oppressive and the wish for sustaining existing reciprocal relationships becomes increasingly weighty. Thus we have that θ(m) is first decreasing in
m and after some point increasing in m. The Er curve will for that reason behave
similarly because of its dependency on θ(m). Figure 4.4 below depicts this case11 .
Er
6
4
Eu
2
0
-2
0
0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9
m2
m3
~
m
1
m
FIGURE 4.4.
An interesting feature of the variable subjective valuation is the possibility
of multiple interior equilibria. For example, in figure 4.4 there are two interior
equilibria, of which one is stable (m3 ). For small market sizes below m2 the search
costs are too high to enter the market. However, at larger market sizes the search
costs are lower and the subjective valuation of the market is higher. This is true for
1 1 Here
{a, b, c, g, h, j, ym , δ} = {.8, .6, 1, −.24, .24, .7, 8, .8} where b(m) = bm and the subjective valution is
given by θ(m) = h + gm + jm2 .
Chapter 4. On the Viability of Gift-exchange in a Market Environment
94
all market sizes between m2 and m3 . Reciprocal exchange is no longer enforceable
and agents enter the market. But contrary to the case in the previous section
where θ is monotonically decreasing and where eventually everybody would be
engaged in market exchange, there is a point m3 at which the market is so large
that agents have relatively high preferences for reciprocal exchange again. At
this point, the market will stop growing and part of the population will stay
in their reciprocal exchange relationship. An interior long-equilibrium solution is
therefore obtained, potentially explaining why gift exchange continues to exist in
the contemporary environment that became more market oriented until now.
4.3.3 Valuation, patience, and viability.
Next, we turn to comparative statics. In general, reciprocal exchange can be
enforceable for some small market sizes but not for large markets. The extent
to which reciprocal exchange is enforceable is first of all obviously depending
on the relative valuation (θ) of reciprocal exchange. Likewise, it depends on the
discount rate (δ) which can alternatively be interpreted as a measure of trust (see
also section 4.2.2). The next two propositions characterize the general relation
between the two variables.
Proposition 5 For every δ ∈ (0, 1) there exist positive θ(δ) such that ∀θ(m) ≥
θ(δ) reciprocal exchange is enforceable and ∀θ(m) < θ(δ) reciprocal exchange is
not enforceable. θ(δ) is first decreasing in δ and then, if θ(m) > θ(0) for some m,
possibly increasing in δ.
Proposition 6 For every δ ∈ (0, 1) there exist positive θ(δ) such that ∀θ(m) ≥
θ(δ) market exchange is not enforceable and ∀θ(m) < θ(δ) market exchange is
enforceable. θ(δ) is first decreasing in δ and then possibly increasing in δ.
Proof. All proofs of the propositions appear in the appendix.
Based on these propositions, figure 4.5 represents the typical shape of the θ-curve
and the θ-curve for all possible combinations of θ and δ.
In this figure, we plotted the areas for which reciprocal exchange is not enforceable for any market size, for which it is enforceable at positive market sizes,
and for which it is enforceable at any market size. The latter implies that market
exchange is not enforceable at any market size. Thus for example, at e1 reciprocal
exchange is not enforceable at any market size (θ < θ), whereas at e3 market ex-
Equilibrium
95
2
• e3
2
• e2
• e1
2
*
0
FIGURE 4.5.
change is not enforceable at any market size (θ > θ). The point e2 is an in-between
case where at some market sizes, but not all, reciprocal exchange is enforceable,
and at some market sizes market exchange is enforceable. The latter case is exactly the one which figures 4.2 and 4.3 are based on. In contrast, e3 for instance
would describe the case where the Eu curve lies entirely above the Er curve.
As can be seen from Figure 4.5, at higher discount factors reciprocal exchange
is enforceable at lower values of θ. In other words, when the discount factor or
the measure of trust is low, cheating is relatively profitable and the subjective
valuation of the reciprocal good must consequently be high for reciprocal exchange
to be enforceable. Exactly the reverse is true for the case of market exchange,
i.e. at higher discount factors, it becomes less likely that market exchange is
enforceable12 .
Here we have a resemblance with optimal contract theory. For example in
Baker et al. [1994] a firm has to choose an optimal bonus system. They can
either rely on an objective but imperfect performance measure or on an unbiased
but not objectively measurable variable. In the former case the firm can rely
1 2 The
interpretation that for high discount rates the θ and θ curves can be increasing is somewhat complicated
and is deferred to the appendix.
Chapter 4. On the Viability of Gift-exchange in a Market Environment
96
on an explicit contract (we loosely interpret this as the money reward in case
of the market exchange), in the latter only on an implicit contract (i.e. based
on trust). They show that under appropriate conditions, the optimal contract
is a combination of the implicit and the explicit contract. But whenever the
discount rate is sufficiently high only implicit contracts should be used whereas
for sufficiently low discount rates one should rely on explicit contracts only.
4.4 Welfare
In this section we take a welfare perspective by comparing the efficiency of stable
equilibria. We follow Kranton [1996a] by taking the weighted discounted lifetime
utility of agents on the market as the measure of comparison, but none of the
results hinge on this. The weights are the shares of the employed and unemployed
agents. Thus, we have:
·
¸
m
m
Vw ≡ 1 −
Vu +
Vm .
(4.10)
1−r
1−r
Evidently, Vu ≤ Vw ≤ Vm . Since both Vu and Vm are increasing in the market
size, so is Vw . We now state:
Proposition 7 It is possible that there exist stable equilibria which are Paretodominated by other equilibria. As a consequence, inefficient market sizes can be
sustained.
Proposition 7 relies on a rather strong criterium of Pareto-optimality, namely
where state i is socially preferred to state j if in state j no agent is worse-off than
in state i and at least one is better off, without taking into account the possibility
of income redistributions. This is stronger than the Kaldor-Hicks criterium and
the Pareto-criterium where income redistribution is allowed. Clearly, under the
weaker versions of Pareto-optimality proposition 7 is as well satisfied.
The rational behind proposition 7 is intuitively clear. The reason that dominated stable equilibria can be maintained is caused by the existence of an external effect and a coordination failure. In essence, if reciprocal exchange is initially
large, search costs on the market are high, even though search costs would be low
if the market was large. Similarly, if the market starts out large, search costs are
low and reciprocal exchange relationship may not be enforceable even though if
Welfare
97
the market was small they would be preferable in value terms. Besides this thickmarket externality there is also a coordination failure. People do not take into
account the full value of a reciprocal trade in deciding between market and reciprocal exchange. A social planner would value reciprocal exchange by its present
value: Vr . But individuals only take into consideration the value of reciprocal
trade for which it can be trusted upon that the relationship can be maintained:
−c + δVr (see definition 1). The gap between those values can be considered as a
coordination failure caused by mutual distrust. As a consequence, if the market
starts out large, there can be a degree of trust that is insufficient to maintain the
reciprocal exchange relationship even though if it could be maintained it would
be superior to market exchange in value terms. Proposition 7 is illustrated below
by means of a graphical treatment (see the appendix for the formal conditions).
Vr
Vw
Eu
Er
mo
~
m
FIGURE 4.6.
In Figure 4.6 the initial market size is denoted by m◦ . At m◦ , definition 1 is
not satisfied, and the economy tends to move to m̃, the maximum sustainable
market size (r = 0)13 . It is immediately seen that at m̃ the value of being in
1 3 It
4.
is evident that the economy could as well converge to a stable interior equilibrium such as m3 in figure
Chapter 4. On the Viability of Gift-exchange in a Market Environment
98
Vw
Vr
Eu
Er
~
m
mo
FIGURE 4.7.
a reciprocal exchange relationship is higher than that of being on the market:
Vr > Vw . Therefore, if all agents would be involved in reciprocal exchange they
would all be better off. Figure 4.7 illustrates the case where reciprocal exchange
is sustainable although everyone could in principle be better off by entering the
market.
The preceding analysis shows that individual agents need not necessarily select
the socially most efficient exchange mechanism nor that there is any tendency
towards a Pareto-optimal equilibrium. There are of course ways to change incentives towards the first best solution, such as a subsidy on entering the market or
lowering search costs in one way or another. Due to the hysteresis present such
a subsidy need only be a temporary one. Stimulation to form reciprocal relationships seems to be of a more difficult order. A one-sided inquiry into the market
mechanism clearly would disguise aspects of critical importance.
4.5 Concluding remarks
Many economic activities imply a certain gift-dimension. In the literature gifts
are discussed from different perspectives (see chapter 2). In this chapter we hold
99
Concluding remarks
the view that the logic of the gift contains some notion of reciprocity. There are
no free gifts in case the institution of gift-giving is considered as a coordination
mechanism to cope with the state of nature. Market exchange and gift exchange
have therefore something in common: the reference to the notion of reciprocity.
But reciprocal exchange is based on personal relationships and trust whereas
market exchange is based on anonymity and money.
Here we assume that market exchange and reciprocal exchange can be made
commensurable by extending preferences. Gift-giving, and thus reciprocal exchange, renders symbolic utility as people value the idea of belonging to a group
or society at large. Symbolic utility and substantive utility, which relates to traditional economic activities, are different components of the extended utility function.
Starting from these premises the main question to be answered is under which
circumstances reciprocal exchange is viable and not crowded-out by the market
regime. Such a form of crowding-out can be conceived as a complete commodification of society. Following Kranton [1996a] reciprocal exchange is modelled
in a strict manner. Agents expect a counter-performance. Market exchange is
modelled as a search process where in case of matching goods are exchanged immediately. The model is closed by introducing an enforceability condition showing
under which conditions agents defect in case of reciprocal exchange. The condition critically depends on the market size and on the discount factor. The latter
has its resemblance in the optimal contract literature once we loosely interpret
market exchange as an explicit contract and a reciprocal relation as an implicit
contract.
It is shown that complete commodification obtains if symbolic utility declines
as the relative size of the market increases. However, reciprocal exchange may
survive for low subjective discount rates which are in some sense indicative for a
high level of trust. There is another reason why full commodification may not be
the equilibrium outcome. Symbolic utility may rise if reciprocity becomes scarce
as the market takes over. In the end people may be aware of a social loss and
revalue reciprocal exchange accordingly. As a result commodification will stop
at some point and there will be an interior equilibrium solution with market
exchange and reciprocal exchange coexisting.
From a welfare point of view, full or partial commodification is not necessarily
superior to reciprocal exchange. It is shown that agents need not necessarily select
Chapter 4. On the Viability of Gift-exchange in a Market Environment
100
the exchange mechanism that is Pareto-optimal. This result is related to the
search externality in market exchange and the coordination failure in reciprocal
exchange. If the market is large initially, search costs are low and the economy
may converge to an inefficient outcome. Similarly, the economy may converge to
an inefficient outcome if the coordination failure with respect to the choice of
reciprocal exchange is important.
What seems most urgent in additional research is a more flexible way of modelling reciprocal exchange to cope with the different aspects of gift-giving. In
particular, it may be rewarding to shed some light on intergenerational gifts,
where reciprocity in the usual sense is out of reach or where the time interval is
extremely long. The latter is for example the case in family relationships where
parents take care of their children in the hope of mutual care once they themselves
become dependent on their children’s readiness to support them. In addition it
may well be more realistic to assume that some goods are more suited to be produced on the market (e.g. bread), others to be offered in a reciprocal relationship
(e.g. baby-sitting), or in a combination of these (e.g. insurance, see Arnott and
Stiglitz [1991]. Individuals can be assumed to be heterogeneous and spend their
time in different proportions over the two exchange mechanisms. Another interesting question emanates from the paradox formulated by Etzioni [1988, 250]:
”The more people accept the neoclassical paradigm as a guide for their behavior,
the more the ability to sustain a market economy is undermined”. The paradox
suggest that market exchange becomes less efficient if morality is on the retreat.
This need not be true, as our analysis suggests, but it certainly deserves serious
consideration.
Appendix
101
4.6 Appendix
In this appendix we first show that definition 1 characterizes a Nash-equilibrium
and subsequently proof propositions 5 to 7. Where no confusion can arise we
simply write θ instead of, for example, θ(m).
Proof that definition 1 is a Nash-equilibrium.
Consider the payoff matrix below. Players are denoted by P and C, the agent
that produces and the one that consumes first respectively. They can either be
honest, H, or cheat, D (defect). The pay-offs are in utility terms. As an example,
consider P playing honest and C cheating, then P produces the good, bearing
cost in utility terms of c and enters the market the next period after finding out
of being cheated, δVu . C consumes yr , does not produce in return and enters the
market getting δVu . Thus in the upper-right cell we have the pay-offs −c + δVu
and yr +δVu . Consider then the strategies {H; H}. This can, by construction, only
be a (weak) Nash-equilibrium if it is in both players advantage not to deviate:
−c + δyr + δ 2 Vr ≥ δVu and −c + δVr ≥ Vu . Fortunately, we can show that first
1
1
inequality is implied by the second. Note that (1−δ)
yr ≥ 2(1−δ)
(yr − cr ) = Vr (see
2
equation 4.5). Then δyr + δ Vr ≥ δVr . If now −c + δVr ≥ Vu (second constraint)
then it is surely the case that −cr + δyr + δ 2 Vr ≥ Vu ≥ δVu which proves the first
inequality. Finally note that {D; D} is always a (weak) Nash-equilibrium.
P ↓; C →
H
2
H
−c + δyr + δ Vr ; yr − δc + δ 2 Vr
D
δVu ; δVu
D
−c + δVu ; yr + δVu
δVu ; δVu
Proof of Propositions 5 and 6.
We try to find values of θ for which definition 1 is satisfied with equality: −c +
δVr = Vu . Label this equality D1. For this value agents are indifferent between
reciprocal exchange and entering the market given the market size. For higher
values of θ they prefer reciprocal exchange, for lower values they prefer market
exchange. In particular we try to find values of θ, say θ, for which the equality
is satisfied but will not be satisfied for any lower θ for any market size. It is
instructive first to consider the case where θ is (weakly) monotonically decreasing
in m and then generalize the results, as in the main text. The advantage of this
is that if reciprocal exchange is enforceable at all, it is certainly enforceable at
Chapter 4. On the Viability of Gift-exchange in a Market Environment
102
m = 0 since Vr is decreasing in m and Vu increasing in m. Thus, we can focus on
m = 0. Then D1 reads, by restricting Vu to R+ :
1
δ(θym
2
+ c) − c
= 0.
1−δ
(4.11)
Since we put no restrictions on θ, given any δ ∈ (0, 1) there exists a θ such that
the equality holds (and is in this particular case easy to find). (As a marginal
comment, note that as δ → 1, the numerator of the LHS of (4.11) approaches
1
(θym − c) which equals zero for some 0 < θ ≤ 1.). Since the LHS of D1 is
2
increasing in δ, θ is lower for higher values of δ.
Generalizing the argument, we see that the RHS can be positive for markets
m > 0, but also that θ(m) may be larger than θ(0). So even if the equality
holds at m = 0, it may well be that the LHS is larger than the RHS at positive
market sizes. The critical value of θ (i.e. θ) can therefore be lower than the value
of θ for which the equality holds at m = 0. The caveat in the generalization is,
however, the fact that Vu is now increasing in δ as well. At m = 0, Vu = 0 no
matter what the rate of time preference is. This remains true up to the market
size where Vu becomes strictly positive. For the range of values for which Vu is
strictly positive, Vu is also increasing in δ. Thus both sides can be increasing in δ
at some market sizes, and the relation between θ and δ becomes ambiguous. We
stress however that for low values of δ, Vu = 0, and hence there exists an interval
where θ and δ are negatively correlated. The intuition behind this increasing part
is that because on the market costs are made before revenues, for sufficiently
low discount rate the present value is negative, whilst in a reciprocal exchange
with some probability you consume before you produce and so even for low (but
positive) discount rates expected gains are positive for some valuation ratio.
As a special case, if θ(m) ≤ θ(0) ∀m (in other words θ is nonincreasing), then θ
can be determined by inspection of m = 0 alone (since if then reciprocal exchange
is not enforceable at m, and since Vu is nondecreasing and Vr nonincreasing in
the market size, then it is not enforceable at any m). Since at m = 0, Vu =
0, if the discount rate increases a little, Vu remains zero but Vr increases so θ
unambiguously declines. But note that such an unambiguously declining θ-curve
is only a special case and that it is not directly related to the shape of θ(m).
The same line of argument can be used to derive proposition 6. Here the aim
is finding the θ such that market exchange is just enforceable at one particular
market size, and not for any higher θ. We first try to find θ in the case of non-
Appendix
103
increasing θ so that the focus can be restricted to m = 1. We do not state the
proof here.
Proof of proposition 7.
The proof consists of showing that it need not be contradictory to have a stable
equilibrium that is inefficient in the sense that it is Pareto-dominated by another
stable equilibrium. Denote the two equilibria under investigation by mi and me
(the subscripts stand for inefficient, efficient). Let the initial point be the inefficient equilibrium mi . Three cases are to be considered: 1. mi is at one of the
corners of the economy. If mi = 0 then it is stable if −c + δVr ≥ Vu . 2. If m = m̃
then it is stable if −c + δVr > Vu . 3. mi is an interior solution. It is stable if
dVr
dVu
−c + δVr = Vu and if δ dm
≥ dm
.
i
i
We have to show that the following set of equations need not be inconsistent:
Vr (me ) ≥ Vr (mi ),
(4.12)
Vw (me ) ≥ Vw (mi ).
(4.13)
Additionally, the enforceability constraints have to be satisfied as indicated at mi
and me . Consider for example the case where mi is the corner solution m = 0
and me is an interior stable solution that Pareto-dominates the corner solution.
Thus we have:
Vr (me ) ≥ Vr (0),
(4.14)
Vw (me ) ≥ Vw (0),
(4.15)
−c + δVr (0) ≥ Vu (0).
(4.16)
−c + δVr (me ) = Vu (me)
(4.17)
0
The second inequality is naturally satisfied. (Indeed, since Vw (m) > 0, the only
case where an equilibrium can be dominated by a smaller market size is where
m = 0 since otherwise all remaining market participants would lose some welfare.
Except, of course, when nobody stays). Since we put no restrictions on θ the first
inequality can be satisfied as well (not, however, when θ(m) is nonincreasing in the
market size). Combining the (in)equalities we see that as long as Vr is increasing
over the interval [0, me ] (but remember that it may be decreasing in the first stage
and increasing thereafter), but not as fast as Vu there is no inconsistency and it
cannot be ruled out that m = 0 is indeed inefficient. Other cases can be analyzed
Chapter 4. On the Viability of Gift-exchange in a Market Environment
104
in a similar manner and are omitted. The result of possible inefficiency is easily
extended to the nonweighted case for Vm , Vu , and Vw all behave in a similar way
(namely increasing in the market size). ¥
5
Welfare Gains of Labeling with Heterogeneous
Consumers
Diamonds are a girl’s best friend. But not only her’s: they are no less a rebel’s best
friend. Rebel armies use guns to force labourers to dig holes and earn hundreds
of millions of dollars by selling the ”blood stones”. To stop these activities, many
countries have now signed to support a system of certificates for diamonds without
a conflicting history. When does this lead to welfare gains?1
5.1 Introduction
Implicit in the standard formulation of the fundamental welfare theorems is that
the characteristics of commodities are observable to all market participants (MasColell et al., [1995]). However, in many cases the consumer cannot observe all
characteristics of a specific good, such as the safety or the quality level. Due to
these informational asymmetries, the consumer is not willing to pay price premiums for different goods. No distinct markets can therefore exist for goods that
are differentiated with respect to unobservable characteristics. As is well known,
this can have dramatic consequences for the efficiency of the market mechanism.
Famous in this respect is the market for lemons as described by Akerlof [1970].
0 This chapter is coauthored by Theo van de Klundert. We are thankful to Richard Nahuis and Sjak Smulders
for helpful comments.
1 See The Economist [2003] and BBC News [2000].
Chapter 5. Welfare Gains of Labeling with Heterogeneous Consumers
106
An in particular interesting class of goods where some characteristics are unobservable is that of goods with social externalities of production. For example,
some production methods have damaging effects on the environment, involve
child labour, or rely on what are perceived as unfair wages. These externalities
are not taken into account by the producers and are therefore not reflected in the
consumer price. And even though it seems that many consumers are in principle
willing to pay premiums for the use of production methods that do not, or to
a lesser extent, involve such social externalities, they cannot observe from the
end product which production method has been used. Hence, goods with social
externalities of production fall into the class of asymmetric information. As in the
case of the lemons market, without improving the consumers’ information, there
is little hope that the equilibrium will be efficient.
In this chapter, it is argued that goods with social externalities of production
pose an even more severe problem to the efficiency of markets than most other
goods with unobservable characteristics. In other cases where some characteristics
are not directly observable, producers can often still signal them to the uninformed
party. For example, the price, advertisements or warranties can sometimes provide
a credible signal to the consumer that the product in question is of high quality
(Tirole [1988]). However, such signaling strategies often do not exist for goods
with social externalities of production (see the next section).
Lacking the possibilities of the usual market responses to informational asymmetries, the government can decide to intervene. One obvious way is to impose a
standard on production methods. Under some circumstances this can be welfare
improving. However, during recent years government regulation is increasingly
relying on information provision to alter behavior (Magat and Viscusi [1992]).
One example is labeling. Labels are certificates issued by a third party that provide credible information about the contents of a product. By now there exists a
variety of such labels that concern, among other things, the environment, working
conditions, fair trade, and child labour. Should they wish to do so, consumers can
contribute to a reduction of social externalities by buying these labeled products.
The crucial difference between standards and labels is the fact that labels are
voluntary. Firms can decide whether or not to apply for the label, and adjust their
production technology conform the requirements. In contrast, standards imposed
by the government are mandatory for all firms in the market. Labeling schemes
therefore allow for more flexibility in the choice of production technology. This
107
Background
has the advantage that labels serve the consumers’ needs better than standards
when consumers are heterogeneous in their willingness to pay for reducing the
social externality. The scope of differentiation is however limited by the costs
of passing on informational contents to the consumers. These costs consist of
designing the label, screening costs of the firm made by the third party, and the
costs of information acquisition by the consumers, each of which can be significant.
Given this trade-off between flexibility and costs, the aim of this chapter is to
examine under which conditions labeling is preferred over standards.
The setup is as follows. The next section first reviews background information concerning markets with asymmetric information and some of the aspects
of labeling. In section 5.3, the model is described and a derivation of the welfare
under imperfect information is given, together with the optimal standard and
label technology. Section 5.4 presents the main proposition where a comparison
is made between the welfare level under a standard and under a labeling policy.
The chapter ends with a discussion and conclusion.
5.2 Background
5.2.1 General background
In this chapter the class of goods is considered whose production methods directly
affects the well-being of the consumers in the economy. This may for example
occur because of the emissions it generates, or because the consumers are altruistic
towards other people who are in some way affected by the production (emissions
in their neighborhood, the wage or working conditions of involved employees
etc.). This class of goods will be referred to as goods with social externalities of
production, in short social externalities.
It is now being recognized that some consumers are willing to pay a price premium for goods whose production methods involve less social externalities (see
for example Kirchhoff [2000]). Their motivations to pay more can stem from as
diverse reasons as a ’warm glow’ feeling from donating (Andreoni [1989]), a sense
of equity (Fehr and Schmidt [1999]), a feeling of guilt or responsibility (Frank
[1988]), or the bestowal of social approval (Holländer [1990], chapter 3). Despite
the willingness to pay a premium, however, producers are discouraged from supplying goods with less social externalities. As pointed out in the introduction,
consumers cannot distinguish by inspection from the end-products which pro-
Chapter 5. Welfare Gains of Labeling with Heterogeneous Consumers
108
duction method has been used. A producer with a more expensive production
method that generates less externalities has to charge a higher price for his product in order not to make losses. The higher price is however not always a credible
signal to consumers of a more expensive production method: producers with a
cheaper production method can raise their price and pretend to have an expensive
production method as well. Clearly, if the low cost producers have an incentive
to imitate the high costs producers, this cannot be an equilibrium.
The above arguments are reminiscent to markets where different producers
offer a variety of quality levels but the quality difference is unobservable to the
consumers. The literature on industrial organization has studied in detail when
and how it would still be possible for producers of high quality to give a signal
to the consumers that their quality is high. The conditions under which this is
possible depend on the type of goods under consideration.
In Tirole [1988] a distinction is made between search goods, experience goods,
and credence goods. In case of search goods, the consumer can determine the
quality in advance by simple inspection. Here, no signalling is needed at all.
With experience goods, the quality is only revealed after consumption. Possible
signalling strategies include advertisements and introductory prices (see Milgrom
and Roberts [1986]). The basic mechanism is that consumers will return to the
producer if they find out that the quality is good. A producer of high quality goods
can then, for instance, give low introductory prices, knowing that the consumers
will return, and reap the lost profits back later. A producer of low quality goods
cannot imitate this strategy, since low introductory prices mean lower shortrun profits and this loss will not be made up in later periods. In effect, a low
introductory price signals that the quality must be high. Finally, credence goods
offer almost no opportunity to learn the quality, even not after using the product
for a long time. In this case, signalling strategies break down because consumers
stay uninformed about the quality even after consuming the good. Hence, there
is no particular reason why consumers should return to this particular producer,
and the strategy to give up short run gains in exchange for long run gains loses
credibility.
The class of goods that is considered in this chapter belongs to the latter group
of goods, in essence credence goods, since the consumer cannot learn anything
about the production method either on beforehand nor by experience, a view
shared for example by Kirchhoff (2000). Labeling may in this case be a good
109
Background
alternative to communicate the information of the production method. We now
turn to a brief overview of the practice of labeling.
5.2.2 Labeling
Labeling provides information. A label can be a statement about the production
method of a good. For example, it can state that fair wages are paid or that no
CFKs are used. Another possibility is that labels are “trademarks” that are associated with certain contents, like some so called ’fair trade’ labels are associated
with fair wages. Of course, labeling is a very crude way of passing on information.
For instance, the statement that ‘fair wages’ are paid leaves a lot of discretion to
the actual wages that are paid. However, the contents of such a message can be
subject to legal restrictions.
In principle, labels do not have to be different from brands, like in the case of
the fair trade label. However, when a lot of small firms are active on the market, they may want to signal their environmental quality collectively to target a
broader public. One way to achieve this is to participate in collective labeling, separated from the individual brands. Collective labeling can also be demand-driven.
Consumers have to invest time in studying the labels, as they have cognitive limitations in absorbing information. Not only do they have to recognize the labels,
they also have to understand what they say, for example what it means that
no CFKs are used. Several studies show that the limitation on human cognition to process information from labels is significant. For example, in one study
concerning labels containing hazard warnings, it is found that as the amount of
information is increased, the consumer’s recall of other information on the product’s label declines, and furthermore that label clutter easily leads to problems of
information overload (Magat and Viscusi [1992]). Naturally, consumers will only
incur these learning costs when these are outweighed by the benefits, which is
more likely when there are only few labels. This favours collective labeling.
Labeling is voluntary. An important feature of labels is that firms voluntarily
participate. A standard, on the other hand, is enforced by the government. The
crucial difference therefore is that under a labeling scheme, some firms may decide
not to participate at all, to serve that part of the market which values quality less.
With minimum standards all firms are required to meet the specified standard.
This latter observation gains importance when consumers are heterogeneous in
their willingness to pay.
Chapter 5. Welfare Gains of Labeling with Heterogeneous Consumers
110
Certification is performed by a third party. The intervention of a third party,
possibly the government, ensures credibility. Consumers that buy labeled goods
must be able to trust the certificate otherwise they buy non-labeled goods.
Examples. There are numerous of instances where labels are used. Most labels
concern the environment, working conditions or fair trade. Some of the more well
known examples include the German Blue Angel and Scandinavian Nordic Swan
eco-labels, and the international Fair-trade label, to name just a few. Another
more recent attempt is the Kimberley process, which has as purpose to certify
diamonds. The latter label is an attempt to prevent rebel armies from using
revenues to buy weapons (The Economist [2003]).
5.3 Description of the model
This section describes the general features of the model.
Producers. The focus is on goods that have social externalities of production.
There are many producers competing on a market of perfect competition. Each
producer can choose a production technology. The available production methods
are characterized by a ’social responsibility index’ s. One may think of s as
representing investments in green technologies or the wage level paid to employees.
Henceforth we refer for simplicity to s as the quality level. It is supposed that
s belongs to [s, ∞). A higher s is interpreted as more investments in reducing
negative social externalities tied to production and is therefore more costly in
terms of production.
The unit costs of producing a good of quality s is denoted by c(s) and this
function is assumed to be convex. In order to derive closed form solutions, it is
assumed that the unit cost is quadratic in quality:
c(s) = αs2 ,
α > 0.
(5.1)
The timing of the game will be that firms first simultaneously choose their quality
levels and then they simultaneously set their prices.
Consumers. As pointed out in the previous section, consumers may care about
the social externality because it directly concerns them (as with effects on the
environment) or because they are altruistic towards the victim (as with child
labour and underpaid workers). To model this, we consider a vertically differentiated product space where each consumer consumes either one or zero units of
Description of the model
111
the good. Let the consumers’ preferences be given by:
(
U = θs − p
if he buys at a quality level of s at price p,
U =0
if he does not buy.
(5.2)
In this case, a higher θ implies a higher willingness to pay for a higher quality,
which represents more concern for the production method. This can be reasonably
attributed to a higher income level but surely also to other exogenous factors, such
as education.
Consumers are heterogeneous in the sense that they have a different willingness to pay to reduce social externalities, e.g. because their incomes differ. The
parameter θ appears to distinguish between different types of consumers. It is
assumed that θ is uniformly distributed on [θ, θ̄] with density f (θ), c.d.f. F (θ),
and θ ≥ 0. Without loss of generality, assume that their total mass is unity. Also
define θa to be the arithmetic average type: θ a ≡ (θ + θ̄)/2. Furthermore, to cut
down on the many possible cases and in order to focus on the interesting ones,
the following assumption on the taste parameter is made:
Assumption 1 θ > αs.
This assumptions states in effect that when the lowest possible quality is supplied
at its competitive price, the market is covered, in the sense that even the lowest
type is willing to purchase a good on the market.
Welfare. For the comparative statics we rely on interpersonal comparable utilities
and define welfare as the sum of total consumer and producer surpluses. By the
assumption of perfect competition, no producer surplus exists in equilibrium.
Hence:
W =
Zθ̄
U(s, θ)dF (θ).
(5.3)
θ
We now examine the welfare level under different policies.
5.3.1 Imperfect information
As a first benchmark, consider the case where consumers have no way to determine
the quality of the product. As the firms cannot discriminate themselves from each
other, it will be clear that there can only be one price in equilibrium. Furthermore,
Chapter 5. Welfare Gains of Labeling with Heterogeneous Consumers
112
competition implies that this price equals the marginal cost of supplying at quality
s: p(s) = c(s). Moreover, it will be clear that given the demand functions in the
second stage, producers have an incentive to cut on quality in the first stage.
Consumers foresee this and (rightly) expect the minimum possible quality level
s.
Note that assumption 1 implies that θ s > αs2 = p(s), so that under imperfect
information the market is covered. This leads to the following proposition:
Proposition 8 In the imperfect information equilibrium all firms supply a good
of quality s at price p = αs2 . Total welfare is given by W I = θa s − αs2 .
Proof. All proofs are in the Appendix.
5.3.2 Standards
Suppose next that the government wants to improve on the imperfect information
equilibrium by imposing standards, i.e. the government requires the quality to be
at least σ ∈ [s, ∞).
It is assumed that the consumers are aware of the standard. As in the case of
imperfect information, consumers also rightly foresee that producers will supply
the good at the minimum quality allowed, that is σ. Competition leads again to
marginal cost pricing.
The consumer who is indifferent between buying or not has a taste parameter
θ̃ such that θ̃σ = p = c(σ). All consumers with a taste parameter equal or greater
than θ̃ = θ̃(σ) buy the good, all the others do not. If the government wants to
maximize welfare, it should therefore choose the standard according to:
σ∗ ∈ arg max W s(σ) =
σ
Zθ̄
(θσ − c(σ))dF (θ).
(5.4)
θ̃(σ)
The first-order condition to this problem is, by Leibniz’s rule, given by:
Zθ̄
∂
∂
(θσ − c(σ))dθ − (θ̃σ − c(σ)) θ̃(σ) ≤ 0,
∂σ
∂σ
(5.5)
θ̃(σ)
with equality if σ > s. The second term drops out since this is the utility of the
consumer who is indifferent between buying or not buying and he has zero utility
Description of the model
113
by definition. When integrated out, one can write the first order condition as:
·
¸
1
(θ̄ − θ̃) (θ̄ + θ̃) − 2ασ ≤ 0,
2
(5.6)
with equality if σ > s. It can be shown that at the optimum the indifferent
consumer is not the highest type: θ̃ 6= θ̄. The intuition behind this is that no
standard should be chosen such that no type consumes. This would give a welfare
equal to zero, whereas welfare would be strictly positive for a standard equal to
s. Hence, the term in brackets must be equal to zero in equilibrium. If θ̃ > θ then
equation (5.6) has solution σ∗ = θ̄/(3α). This, implies that θ̃ = θ̄/3 which is only
compatible with θ̃ > θ if θ̄ ≥ (3/2)θa . For all other values of θ̄, it must be that
θ̃ = θ (covered market) and σ ∗ = (θ̄ + θ)/(4α).
Proposition 9 Suppose the government imposes a minimum standard σ on s.
Then, if θ̄ < (3/2)θa the optimal standard is given by σ ∗ = max {s, θa /(2α)} and if
©
ª
θ̄ ≥ (3/2)θa then the optimal standard is given by σ ∗ = max s, θ̄/ (3α) . Welfare
3
is equal to to W s = (θa )2 /4α for σ ∗ = θ a / (2α) and W s = θ̄ /(27α(θ̄ − θa )) for
σ∗ = θ̄/(3α).
For low values of θ̄ all consumers consume the standard. The optimal standard
and welfare are in this case only dependent on the average taste parameter θa .
An equal increase in θ̄ and decrease in θ, hence keeping the average constant, has
thus no effect on the optimal standard or welfare. For higher values of θ̄, it does
not pay off to make the low types consume. The optimal standard is increasing
in θ̄ because this means that the high types are willing to pay more.
Notice that social welfare with optimal standards is always weakly higher than
under imperfect information. This must be true since the government can always
reach the same welfare level of imperfect information by setting the minimum
standard equal to s. However, in many case can the government do strictly better
than under imperfect information by increasing the minimum standard above
what is minimally feasible for producers. Note however, that a minimum standard
can be, but is not always a Pareto-improvement upon the imperfect information
equilibrium. The consumer of type θ0 is indifferent between the minimum quality
and a standard if θ0 s − αs2 = θ0 σ − ασ 2 or, provided that σ > s:
θ0 = α(s + σ),
(5.7)
Chapter 5. Welfare Gains of Labeling with Heterogeneous Consumers
114
and trivially at σ = s. The minimum standard σ lowers welfare for all types
θ < θ0 . In particular, if θ < θ0 then some consumers are worse off under the
minimum standard. Their consumer surplus decreases because they now pay a
higher price that does not outweigh the quality increase for them. Some of them,
namely for whom θ < ασ, even stop consuming.
Lemma 2 The introduction of a minimum standard is a Pareto-improvement if
and only if σ ≤ (θ/α) − s.
It is easy to show that a higher s or α create less opportunities for Paretoimprovements. To relate Pareto-improvements to heterogeneity consider the following measure of heterogeneity:
Definition 2 Consumers are more heterogeneous if the taste parameter of the
highest type, θ̄, is higher and the average taste parameter, θa , is kept fixed.
Thus, consumers are more heterogeneous if the spread of types (θ̄ − θ) increases
keeping the average constant. With this definition, the the following proposition
is obtained:
Proposition 10 The optimal minimum standard is a Pareto-improvement if and
only if consumers are not too heterogeneous in their taste parameter, that is, if:
θ̄ ≤ 32 θa − 2αs.
It follows that if consumers are ’sufficiently’ heterogeneous, the optimal minimum
standard benefits the high type consumers and hurts the low type consumers.
5.3.3 Labeling
The latter proposition is interesting, because it lays bare the limits of standards:
with a minimum standard, all consumers are forced to buy at or above the minimum standard, or refrain from buying, including those who have no specific
interest in buying high quality. In this section we discuss another mechanism
mentioned in the previous section that also provides information about the quality but leaves the option of supplying at low quality: labeling.
Characteristic about labeling is that it is voluntary: each firm can freely decide whether or not it wants to carry a certain label and conform to the label’s
specified standards. The government (or another third party) will set up controls
to guarantee that certified firms keep up to the standards in agreement with the
Description of the model
115
label’s content. Consumers, on their side, can freely choose whether or not to purchase a labeled product. If they decide to buy a labeled product, they can read
the quality from the label’s description. If they purchase a non-labeled product,
they infer that it must be of the lowest possible quality.
We already pointed out that there are several cost factors specific to a labeling
policy. First, consumers are limited in their cognitive abilities and therefore have
to make some costs in order to process the information carried on the label. Second, producers may have to restructure their production process, design a label,
and draw public attention to their labeling policy through advertising campaigns
and social responsibility reports. Of course, in order to ensure credibility of the
label’s contents, the certifying third party has to involve in costly monitoring of
the firms’ production process. Such monitoring costs are equally likely for the
case of standards though.
It may be assumed that the costs of informing consumers in particular are
relevant and for the current purpose we therefore restrict attention to these. We
limit the number of possible labels to one and assume that the additional costs,
b, over normal production costs for carrying a label take the following form:
b(τ ) = β(τ − s),
(5.8)
where τ is the quality of the label as specified in the certificate. Hence, it is
assumed that no costs are made for a label that specifies the same technology
as the minimum possible technology (b(s) = 0) and furthermore that costs are
increasing in the quality level. This positive relationships between the cost and
quality of a label may for example be because a higher quality level means a more
complicated technology making it more difficult to well inform the consumer. How
accurate this specification of costs of information exactly is, is unclear, but this
specification helps keeping the analysis tractable and does not affect the main
results in an important way.
Suppose therefore that the government introduces a label for products that are
at least of quality τ . Because of the marginal cost pricing by firms, the price of a
labeled good will be equal to:
p(τ ) = c(τ ) + b (τ ) .
(5.9)
Chapter 5. Welfare Gains of Labeling with Heterogeneous Consumers
116
The type who is indifferent between buying the labeled and the non-labeled product is implicitly defined by b
θs − αs2 ≡ b
θτ − ατ 2 − β(τ − s), or, provided τ > s:
θ̂ = α(τ + s) + β.
(5.10)
No types refrain from consumption in this case: all types θ < b
θ consume the nonlabeled product of quality s, whilst all types θ ≥ b
θ consume the labeled product
of quality τ . Hence, welfare W l is given by:
l
W (τ ) =
Z
θ
b
θ(τ )
2
(θs − αs )dF (θ) +
Z
θ
b
θ(τ )
(θτ − ατ 2 − β(τ − s))dF (θ).
(5.11)
The government then sets the quality of the label as to maximize welfare: τ ∗ ∈
arg max W (τ ). The first order condition is given by:
∂W l
=
∂τ
Z
θ
θ̂(τ )
(θ − 2ατ − β))dF (θ) ≤ 0,
(5.12)
with equality if σ > s. (Note that by the rule of Leibniz one should also take the
changes in the limits of the integral into account. However, this is a change in the
marginal consumer who is by definition indifferent, and hence these terms cancel
out.) To rule out uninteresting cases, we make an additional assumption:
Assumption 2 θa ≥ 2αs + β.
This assumption is sufficient to rule out that the optimal quality of the label will
be so high that no consumer finds it interesting to consume the label. Solving the
first order condition leads us then to the following proposition:
Proposition 11 Suppose the government certifies producers who produce at least
a quality of τ . Then the optimal quality required for a label is given by τ ∗ =
max{s, (θa − β)/ (2α)} if θ̄ < (3/2)θa − αs − (1/2)β and τ ∗ = max{s, (θ + αs −
β)/ (3α)} if θ̄ ≥ (3/2)θa − αs − (1/2)β.
Like with a standard, if θ̄ is relatively low all types consume the labeled product
and the optimal label is independent of θ̄. The optimal quality of the label is
decreasing in the cost parameter β. This is so because when the cost parameter β
increases, by (5.8) the costs of a label increases in quality and so a lower quality
is preferred. When θ̄ increases further it is beneficial to increase the quality of the
117
Description of the model
label. High types are willing to pay a lot while low types can still enjoy a quality
level of s. The optimal quality is again decreasing in costs β, and, similar to the
case for standards, increasing in θ̄. Furthermore, if the cost parameter α is higher
or the minimum available quality is higher, so is the optimal quality. A higher α
or s makes the nonlabeled product more attractive for low types. Now that more
types will prefer the low quality, the high types can be made a bit better off by
increasing the quality of the label.
The expression of the welfare is a bit cumbersome in this case, and is therefore suppressed. But note again that the introduction of a label always weakly
increases welfare. However, contrary to imposing a standard, a labeling scheme is
always a Pareto-improvement upon the imperfect information equilibrium. This
is true, because consumers still have the option of consuming a good of quality s
but their choice is enriched with products that supply quality τ ∗ .
Before coming to the main proposition, it is interesting to discuss how the
optimal label relates to the optimal standard. First note that when everybody
consumes the label or standard, the optimal label quality, (θa − β)/ (2α) , is lower
than the optimal standard, (θa )/ (2α). This makes sense: since the costs of a label
are increasing in the quality, the optimal quality is lower if the costs are higher.
For zero costs (β = 0) the two cases are identical.
Secondly, when not everybody would consume the label or standard, the optimal label is again lower than the optimal standard if the costs are high, but
can be higher if the cost parameter α is high enough. The intuition is that if α
is high, a high standard is bad because everybody pays high costs. But with a
label, only the higher types will consume the label and these are the ones that
are willing to pay most, making it attractive to set the quality relatively high.
Finally, note that the threshold value of θ̄ for which not everybody consumes
the label or standard is lower under the labeling policy, where this threshold
is given by (3/2)θ a − αs − (1/2)β, than under standards, where the threshold
is (3/2)θa . With a fixed average θa , a higher θ̄ means a lower corresponding θ.
Increasing the quality of a label means that low types will consume s. Increasing
the standard means pushing the low types out of the market. Pushing consumers
out of the market is worse than making them consume a nonlabeled product,
which should therefore be prevented. Hence, the standard is chosen such that low
types stay longer in the market.
Chapter 5. Welfare Gains of Labeling with Heterogeneous Consumers
118
5.4 Standards or labels?
We are now ready to formalize the idea mentioned in the introduction that the
trade-off between a standard and a label is that of costs versus flexibility in the
producers’ choices of technology. One the hand brings the label a cost to the
consumer in terms of a higher price. It is clear that if the costs parameter β is
very high, a standard is to be preferred over a label, everything else equal. Also,
when the cost parameter is very low, a label is unambiguously to be preferred.
Indeed, if β = 0, then the welfare under a label is weakly higher than under
a standard, since the government can set τ = σ∗ in which all consumers who
want to can achieve the same utility level under a labeling scheme than with
standards, and some can achieve higher utility by consuming quality s instead of
the standard (or refrain from consuming). For a particular range of heterogeneity,
the welfare of labeling is in this case strictly higher.
The other part of the trade-off is that labeling is more flexible in providing
quality in the customers’ needs. Intuitively, labeling leads to higher welfare than
a standard if consumers are heterogenous in their taste parameter. This intuition
turns out to be only partly correct. In a few steps, we show the following result:
labeling is better than standards only for an interval of heterogeneity. For either
small heterogeneity or high heterogeneity, a standard is to be preferred.
First, for notational convenience we introduce some more notation:
Notation 1 η ≡ (3/2)θa − αs − (1/2)β.
The parameter η is the threshold for which all types are willing to buy the label.
Consider then the following result:
Lemma 3 If θ̄ < η then W l ≤ W s with strict inequality for β > 0.
If θ̄ < η, the optimal standard and label would be such that all types would
consume the standard or the labeled product respectively. In essence, no consumer would refrain from consuming or buy the unlabeled good. The intuition
is straightforward. If β = 0, the optimal standard and label coincide: σ ∗ = τ ∗ .
Clearly welfare would be identical. If the cost parameter β increases, but still all
consumers buy the label, the welfare under a labeling policy must necessarily be
lower. They still prefer the label to s, but the costs decrease their utility.
Next, consider:
119
Standards or labels?
Lemma 4 If η ≤ θ̄ < (3/2)θa and β = 0 then W l > W s .
For values of θ̄ within this range, the welfare of standards is independent of
heterogeneity and that of labeling is increasing in heterogeneity. As W l = W s at
θ̄ = η for β = 0 this lemma follows straightforwardly. As a final intermediate step
assume that θ̄ ≥ (3/2)θa and consider:
Lemma 5 Let ∆W (θ̄, β) ≡ W l − W s . Then under assumptions 1 and 2, and
with θa fixed, ∆Wθ̄ (θ̄, β) ≥ 0 for θ̄ ≤ ξ and ∆Wθ̄ (θ̄, β) < 0 for θ̄ > ξ, where
√ p
ξ ≡ θa + 13 3 (3θa (θa − 2αs − β) + (2αs + β)2 .
(A subscript denotes a partial derivative.) Thus, up to a certain threshold level
of θ̄, ξ, the welfare of labeling is increasing faster in heterogeneity than that of
standards. Beyond this threshold level, the reverse holds. Finally note that W l is
always decreasing in β. Putting things together, the following result is obtained:
Proposition 12 If heterogeneity is small, welfare under standards is always weakly
higher than under a label and strictly higher for any β > 0. For intermediate heterogeneity there exists an interval where labels outperform standards provided that
the cost parameter β is not too high. For high heterogeneity standards outperform
labels again if β is high enough.
Although not all the thresholds levels for which labels are welfare improving compared to standards have been derived explicitly, the picture is clear in qualitative
terms. For intermediate levels of heterogeneity labels are better than standards.
The size of this interval increases as the cost parameter β decreases.
The result of proposition 12 is pictured in the figure below. The figure is drawn
for a positive cost parameter β. Up to the point η, welfare of standards and
labeling are both constant in heterogeneity, though welfare of labeling is lower.
After this point, the welfare of labeling starts to increase and surpasses that of
standards. The difference in welfare ∆W ≡ W l − W s starts to increase. After the
point ξ the difference decreases. The curve shifts upward for lower values of β
and would be positive or zero over the whole domain for β = 0. The curve shifts
downward for higher values of β, and the domain where ∆W is positive ceases to
exist for high enough values of β.
The result that standards are better for sufficiently high heterogeneity is perhaps somewhat counterintuitive and deserves some more attention. Under a minimum standard scheme, consumers have the choice between buying the product
Chapter 5. Welfare Gains of Labeling with Heterogeneous Consumers
120
FIGURE 5.1.
at the quality of the standard, or refrain from buying. The consumer who is indifferent is characterized by θ̃. Under a labeling scheme, the choice is between
a labeled and a nonlabeled product. The consumer who is indifferent is characterized by θ̂. Consider any θ̄ ≥ ξ. Under that condition, θ̃ < θ̂. It follows that
(θ̄ − θ̃) > (θ̄ − b
θ). In words: the fraction of consumers who consumes the highest
available technology is greater under a standard (θ̄ − θ̃) than under a labeling
policy (θ̄ − b
θ). This is intuitive: under a labeling scheme the alternative is a nonlabeled good which is still better than the alternative under a minimum standard
(refraining from buying). Thus, a higher fraction will consume the highest quality under a minimum standard because the alternative is worse. Then, when the
heterogeneity increases, both the optimal standard and the optimal label increase
at the same rate 1/(3α). Under the standard, however, a larger fraction of the
consumer benefits from this increase. This effect is becoming more important for
higher values of θ̄, when heterogeneity increases.
5.5 Discussion and conclusions
In this chapter we have formalized some of the basic trade-offs between labeling
and standards. Conditions under which labeling is welfare improving are derived.
121
Discussion and conclusions
We now discuss the importance of the main underlying assumptions, and the
consequences of relaxing them.
The key assumption underlying the model is that some consumers are willing
to pay a price premium for goods that involve less social externalities of production. This is a necessary requirement for labeling to be an effective instrument.
Although there is evidence that consumers are indeed willing to pay a price premium, the evidence is still somewhat mixed (Bjørner et al.[2002]).
Assumption 1 is a relatively innocent assumption. Allowing for uncovered markets only changes the baseline welfare level W I but qualitatively speaking it does
not affect any of the conclusions.
Assumption 2 ensures that there is a scope for government intervention in the
first place. If assumption 2 does not hold, then the optimal label is such that all
consumers buy the unlabeled good. In this case, no intervention is needed other
than possibly a standard. Hence, this assumptions is a first check for the relevance
of policy analysis.
We end the chapter by suggesting extensions to the model that seem necessary
for a well founded policy and that we intend to incorporate in future studies.
First of all, goods with social externalities of production often have a public
good character. Next to say a feeling of guilt of buying these products, they may
also care about the externality per se. The case of polluting production methods
belongs to this category. All people may care about a cleaner environment, but
there are clear incentives for free-riding. Even if a group of consumers is willing
to contribute to reduce the externality by buying labeled goods, there will likely
also be a group of consumers who prefer the nonlabeled goods. The latter group
imposes a negative externality on society. When the group of nonlabeled product
buyers is large, a labeling policy is hardly effective in reducing the externality.
Standards circumvent this problem as it does not allow for consumers who wish
not to contribute.
Second, it is interesting to take note of the results of chapter 7 that deals
with self-serving biases in information procession. In the current chapter we have
assumed that some consumers may for instance experience a feeling of social
irresponsibility when they buy a good with high social externalities, inducing a
feeling of guilt. It is thus argued that they are willing to pay a price premium to
avoid any feelings of guilt.
Chapter 5. Welfare Gains of Labeling with Heterogeneous Consumers
122
If this is related to the results of chapter 7, it becomes clear that consumers
may nevertheless be motivated to stay ignorant about labeled products. It is
clearly best for the consumers to buy a cheap good without reflecting on the negative social externalities that the product brings along. There is ample evidence
from psychology that many people indeed process information in a biased way.
Moreover, this seems to be mostly going on when their self-concept is at stake,
such as when people behave in a way that does not support their view of being
decent (Aronson [1988]). This is also likely to be the case for some of the products considered in this chapter. The taste parameter itself is then a function of
efforts in gathering information, which on its turn is depending on the costs of
acquiring information. The more information is gathered about the wrongdoing
of a production method, the higher the willingness to pay for a product with a
higher social responsibility index. A labeling policy may well be very effective as
it raises the costs of ignoring the information much more than, say, a tax would.
A second relation with chapter 7 is the following. In this chapter, it is analyzed
which policy instrument is better given the tastes of people. Thus, the feelings
of guilt make a labeling policy possible. Chapter 7 on the other hand, learns us
something how to foster such feelings because it endogenizes the willingness to
pay in relation to subsidies.
Third, the assumption of perfect competition may be relaxed to the case where
firms have some market power. This can have interesting consequences. Take for
example the case where there are only two firms. Under imperfect information and
a minimum standard, Bertrand price competition drives profits to zero. However,
if one firm supplies a labeled product and the other a non-labeled product, then
their products are differentiated and they exercise some market power (for a more
elaborate treatment see Bansal [2002]). In this way, labeling has a negative effect
on efficiency by increasing market power which may or may not be outweighed
by the larger array in the variety of goods.
Fourth, the impact on trade has not been discussed. One common theme in
the discussion on the implementation of labeling is that it imposes hidden trade
barriers (see in particular Keyzer [2002]). If consumers are willing to pay premiums for clean technology, then this may well be in the advantage of producers in
developed countries who have better knowledge about clean technologies.
Finally, it is assumed that the government is perfectly informed about the production technology. However, it may be hard for the government (or any other
123
Discussion and conclusions
certifying party) to obtain precise information on all the stages of the production process, especially if many small producers are involved. Uncertainty on the
governments side can have a great impact on the optimal policy instrument (see
Baumol and Oates [1988]). Furthermore, consumers may become skeptical about
paying price premiums if the contents of a labeled product is open to debate (see
Mason [2002]).
Chapter 5. Welfare Gains of Labeling with Heterogeneous Consumers
124
5.6 Appendix
This appendix contains all the proofs and lemmas in the main text with the
exception of the proofs of propositions 8 and 10 and of lemmas 2 and 4 which
follow from straightforward substitution and/or are elaborated upon in the main
text and are for that reason suppressed.
Proof of proposition 9.
Most of the proof is contained in the main text or is straightforward. To complete
the proof, note that W s is continuously decreasing in σ between the unconstrained
optimum (unconstrained in the sense of ignoring the constraint σ ≥ s) and the
point where θ̃ would equal θ̄, and constant thereafter. Hence, if the unconstrained
optimum is not within the interval [s, ∞) then the constrained optimum is indeed
at s. ¥
Proof of proposition 11.
Integrating out equation (5.12) gives:
(θ̄ − b
θ)((1/2)(θ̄ + b
θ) − β − 2ατ ) = 0,
(5.13)
as the first order condition. If θ̄ = b
θ all consumers purchase the nonlabeled
good and welfare is as under imperfect information. With θ̄ 6= b
θ and using
equation (5.10) we can see that the first order condition (5.13) is satisfied for
θ = (1/3)(θ̄ + 4αs + 2β). Hence,
τ = (1/(3α))(θ̄ + αs − β) and consequently b
b
θ̄ = θ occurs at θ̄ = 2αs + β. For any smaller θ̄ the optimal quality would be
such that even the highest type would consume the nonlabeled product but this
case is ruled out by assumption 1. With the foregoing expressions we can also
see that θ = b
θ occurs at θ = 32 θa − αs − 12 β. In this case, substitution of θ = b
θ
a
into equation gives the optimal quality τ = (θ − β)/ (2α) . These are the unconstrained optimal qualities. Since W l is continuously decreasing in τ between
the unconstrained optimum (with respect to s) and constant after b
θ = θ̄, the
constrained optimum is indeed s if the unconstrained optimal τ is not element of
[s, ∞). ¥
Proof of lemma 3.
For θ̄ ≤ η, W s = (θa )2 /(4α). It is straightforward to show that in this case W l =
(1/2)τ ∗ (θ̄+θ)−ατ ∗2 −β(τ ∗ −s). This can be rewritten as W l = (θa −β)2 /(4α)+βs.
125
Appendix
For β = 0, W l = W s . For any β > 0, W l < W s since by assumption 2 it must be
the case that 2θa > 4αs + β. ¥
Proof of lemma 5.
1. θ̄ ≤ 3θ ⇔ θ̄ ≤ (3/2)θa . First note that it follows immediately from proposition
9 that W s is independent of both β and θ̄. It is also straightforward to show that
W l is decreasing in β. We continue by proving that W l is improving in θ̄.
All types θ ∈ [θ, b
θ) consume the unlabeled product of quality s, all types
θ ∈ [b
θ, θ] consume the labeled product of quality τ . Hence, the welfare under the
optimal label τ ∗ is given by:
Z bθ
Z θ
1
1
l
2
W =
(θs − αs )dθ +
(θτ ∗ − α (τ ∗ )2 − β(τ ∗ − s))dθ. (5.14)
b
θ θ−θ
θ θ−θ
We examine the behavior of W l around the optimum with respect to changes in
θ̄. By the implicit function theorem we have (ignoring other parameters):
¯
dW l (θ, τ (θ)) ¯¯
∂W l (·)
=
.
(5.15)
¯ ∗
dθ
∂θ
τ =τ
In the rest of the proof we suppress the condition that τ = τ ∗ but it is assumed
to hold throughout. Taking the partial derivative of W l with respect to θ using
the rule of Leibniz (and recalling that θ = 2θa − θ)gives:
Z bθ
h
i
1
∂b
θ
∂W l
1
2
2
b
=
−
(θs − αs )dθ +
·
θs − αs
(5.16)
∂θ
2(θ − θa )2
∂θ 2(θ − θa )
θ
Z θ
¤
£
1
1
2
+
θs − αs +
−
(θτ − ατ 2 − β(τ − s))dθ
a
a 2
b
2(θ − θ )
2(θ
−
θ
)
θ
¤
£
1
θτ − ατ 2 − β(τ − s)
+
a
2(θ − θ )
h
i
∂b
θ
1
2
b
− ·
θτ
−
ατ
−
β(τ
−
s)
.
∂θ 2(θ − θa )
The second and the last term of the RHS cancel out against each other. Integrating
out results in:
£
¤
∂W l
1
1
1
= −
[b
θ − θ)( s(b
θ + θ) − αs2 )] +
θs − αs2 (5.17)
a 2
a
2
∂θ
2(θ − θ )
2(θ − θ )
1
1
−
[θ − b
θ)( τ (θ + b
θ) − ατ 2 − β(τ − s)]
a 2
2
2(θ − θ )
£
¤
1
+
θτ − ατ 2 − β(τ − s) .
a
2(θ − θ )
Chapter 5. Welfare Gains of Labeling with Heterogeneous Consumers
After some manipulations, this can be rewritten as:
·
¸
1 b
∂W l
τ −s
a
2
=
(θ − θ) + (b
θ − θ )(θ − α(s + τ ) − β) .
∂θ
2(θ − θa )2 2
126
(5.18)
It is useful to proceed as follows. We distinguish between θ < 3θa − 4αs − 2β and
θ ≥ 3θa − 4αs − 2β. The former case implies that b
θ < θa , the latter case that
b
θ > θ a . For each case we determine when b
θ hits the boundary. If b
θ < θa and b
θ hits
b
the boundary, then it must be the case that θ = θ. (It cannot be the case that
b
θ < θa ). Similarly, if b
θ > θa and b
θ hits the boundary, then
θ = θ since θ ≥ θa and b
it must be the case that b
θ = θ. The thresholds are given by θ = 32 θa − αs − 12 β
(which follows from setting b
θ = θ) and θ = 2αs+β (following from setting b
θ = θ).
a
3 a
1
Note that as a consequence of Assumption 2, 3θ −4αs−2β > 2 θ −αs− 2 β. Note
furthermore that Assumption 2 also implies that θ > 2αs + β which rules out the
possibility that b
θ = θ. We are left with the following three different (exhaustive)
cases:
Case (i): θ < 32 θa − αs − 12 β.
In this case, b
θ = θ. Substituting into equation (5.18) gives
·
¸
∂W l
τ −s
1
a
2
=
(θ − θ) + (θ − θ )(θ − α(s + τ ) − β) ,
(5.19)
∂θ
2(θ − θa )2 2
which is equal to:
¤
∂W l
τ −s £
a 2
a
−
θ
)
+
(θ
−
θ
)(θ
−
α(s
+
τ
)
−
β)
(5.20)
=
2(θ
∂θ
2(θ − θa )2
(τ − s) (θ − θa )
=
[2(θ − θa ) + (2θa − θ − α(s + τ ) − β)]
2(θ − θa )2
(θ − θa )
=
[(τ − s) (θ − α(s + τ ) − β)]
2(θ − θa )2
¢ ¡
¢¤
(θ − θa ) £¡
2
2
θτ
−
ατ
−
β(τ
−
s)
−
θs
−
αs
=
2(θ − θa )2
(θ − θa )
=
[U(τ , θ) − U(s, θ)]
2(θ − θa )2
≥ 0.
The last inequality follows by a revealed preference argument, i.e. the fact that
the lowest type (weakly) prefers the label.
Case (ii): 32 θa − αs − 12 β ≤ θ < 3θa − 4αs − 2β.
Appendix
127
¡
¢
For values of θ within this interval, θ̂ = θ + 4αs + 2β /3. Recall also that the
¡
¢
optimal label is in this case given by τ ∗ = θ + αs − β /3α. Before proceeding,
note that the two conditions on θ̄ imply that θ̄ > 2αs + β otherwise the case does
not exist. Substituting θ̂ and τ ∗ into equation (5.18), and expanding all terms
gives:
·
τ −s
∂W l
4 2 4
2
8
8
=
θ − αsθ − βθ − α2 s2 − αβs
a 2 9
9
9
9
9
∂θ
2(θ − θ )
¸
2
6
12
6
− β 2 − θθa + αsθa + βθa .
9
9
9
9
This can also be written as:
µ
µ
·
¶¶¸
3 a
1
∂W l
τ −s
4
=
(θ − 2αs − β) θ −
θ − αs − β
.
2
2
∂θ
2(θ − θa )2 9
(5.21)
(5.22)
By inspection of the assumptions and the note we made before, it is then easily
seen that this expression must be positive.
Case (iii): θ ≥ 3θa − 4αs − 2β.
¡
¢
We now have θ̂ = θ + 4αs + 2β /3 again. This time, rewrite equation (5.18) as:
·
∂W l
1
1
=
(τ − s)(b
θ − θ)2
a 2 2
∂θ
2(θ − θ )
£
¤i
a
2
2
b
+(θ − θ ) (θτ − ατ − β(τ − s) − (θs − αs )
(5.23)
which is equal to:
·
¸
£
¤
1
∂W l
1
a
2
b
b
(τ − s)(θ − θ) + (θ − θ ) U(τ , θ) − U(s, θ) .
=
∂θ
2(θ − θa )2 2
(5.24)
By a revealed preference argument, it is easy to see that it must be the case that
θ. Under the assumptions made we also have
U (τ , θ) − U(s, θ) > 0 since θ > b
a
b
θ − θ > 0. Hence, we conclude that in this case ∂W l /∂θ > 0.
2. θ̄ > 3θ ⇐⇒ θ̄ > (3/2)θa . Note first that it cannot be the cast that θ̃ > θ
and θ̂ = θ. Hence, if the market for standards is not covered, then also not with
labels. This on its turn means that we can substitute for θ̂ in equation (5.18)
giving us equation (5.22) for ∂W l /∂θ. It is easy to show that:
¯
2
2θ (θ̄ − 32 θa )
∂W s ¯¯
=
.
(5.25)
27α(θ̄ − θa )2
∂θ ¯σ=σ∗
Chapter 5. Welfare Gains of Labeling with Heterogeneous Consumers
128
Subtracting this from (5.22) gives us (again suppressing notation that we examine
behavior around the optimum):
·
¶¶
µ
µ
µ
¶¸
∂(W l − W s )
3 a
1
3 a
2
2
= χ (θ̄ − 2αs − β) θ̄ −
θ − αs − β
− θ̄ θ̄ − θ
,
2
2
2
∂ θ̄
(5.26)
a 2
where χ ≡ 2/(27α(θ̄ − θ ) ). This can, after some manipulations, be rewritten as:
¸
·
∂(W l − W s )
3 2
3 a
a
a
2
= −χ(2αs + β) θ̄ − 3θ θ̄ + 3αsθ + βθ − (2αs + β) .
2
2
∂ θ̄
(5.27)
It is then straightforward to show that:
∂(W l − W s)
=0⇔
∂ θ̄
θ̄ = ξ ≡ θa ±
1√ p a a
3 3θ (θ − 2αs − β + (2αs + β)2 .
3
(5.28)
(5.29)
Since θ̄ ≥ θa we only need to consider the positive root. For any θ̄ Q ξ, ∂(W l −
W s)/∂ θ̄ R 0. This proves the lemma. ¥
Proof of proposition 12.
This follows in a straightforward manner from lemmas 3, 4, and 5. ¥
6
Rewards, Self-confidence, and Motivation: The
Hidden Rewards of Rewards
6.1 Introduction
Many real-life situations concern relationships where no complete explicit contracts can be written down. Moreover, many situations do not allow the possibility of implicit contracts either. Incomplete contracts, however, can often give
strong incentives to shirk (Williamson [1985], Fehr and Gächter [2000]). This
raises some questions such as: why are efforts rewarded, and why are they made
in the first place? A particularly intriguing question is why spontaneous rewards
are sometimes given. This latter question is taken up in this chapter by examining
the role of self-confidence and its effect on motivation.
There are several possible channels through which motivation can be stimulated. Most of economic theory is built on the assumption that monetary rewards
motivate agents to make efforts. The channel through which rewards motivate
is straightforward: agents care about money, and hence they are more willing to
make efforts if this increases the probability of payments1 . Social psychologists
have identified another channel through which motivation can be stimulated,
namely through changes in self-confidence. Whenever an agent is more confident
0 This
chapter is coauthored by Anton Souvorov. We thank Jean Tirole for helpful ideas. Part of the research
was done during my stay at GREMAQ at the University of Toulouse 1, that was supported by the ENTER
exchange and a Marie-Curie fellowship.
1 More generally, rewards can give non-monetary benefits — payments in kind, promotions, recognition. What
matters is that rewards directly affect the agent’s utility.
Chapter 6. The Hidden Rewards of Rewards
130
about his ability to succeed, he is more likely to try to undertake a task. Thus,
if one succeeds in making the agent more self-confident, his effort increases and
a successful outcome is more likely.
The primary aim of this chapter is to study both channels in one simple model.
Although both monetary rewards and self-confidence are elements in economic
models, the self-confidence effect has rarely been studied: in most conventional
models with asymmetric information, the agent has full information about his
ability. Consequently, there is no role for the principal to give signals to the
agent. By contrast, in this chapter it is assumed that the agent has incomplete
information about his chances to attain a successful outcome, for example because
he undertakes a task that is new to him. In this case, it is shown that monetary
rewards can give credible signals about these chances and therefore influence the
level of self-confidence.
Interaction effects between rewards and self-confidence are also examined in a
recent paper by Bénabou and Tirole [2002a]. Their focus is on promised bonuses
that are specified in a contract. The focus in this chapter is on discretionary
rewards. Such unexpected bonuses are interesting in their own respect, because
why would someone give a bonus that is not expected or specified in a contract
anyway? A possible answer is that such rewards can increase self confidence. Or,
in the words of Bénabou and Tirole [2002a, 22]:
”Rewards that are discretionary (not contracted for) may well
boost the agent’s self-esteem or intrinsic motivation, because (...) the
worker or child learns from the reward that the task was considered
difficult (and therefore that he is talented), or that the supervisor is
appreciative of, proud of, or cares about his performance — and that
it is worth repeating it. (...) And receiving the reward is good news,
because the agent initially did not know how to interpret his performance.”
The main idea of the model in this chapter is that there is an agent who is
only willing to make efforts if he has enough self-confidence that he will succeed
in the task. However, it is difficult for him to assess the outcome. By giving a
bonus, the principal can give a signal to the agent that his efforts resulted in a
successful outcome last period. Thus, bonuses increase motivation in our model in
two ways: first of all, they raise the agent’s self-confidence, and he realizes that it
Introduction
131
is worthwhile for him to continue working hard; secondly, foreseeing that bonuses
will be given more frequently after successful outcomes, the agent works harder in
the first place. Altogether, this can explain why sometimes ”unexpected” rewards
are given, even in a game with a finite number of rounds2 . These results are in line
with actual behavior in existing markets (Akerlof [1982]) as well as in laboratory
experiments (Fehr and Gächter [2000]).
Some social psychologists have stressed that rewards need not necessarily increase self-confidence. In their view, rewards sometimes have ”hidden costs” (e.g.
Kohn [1993]). The hidden costs of rewards are the possible negative effects on
self-confidence. There are two main causes for this. First, rewards can be perceived as controlling, thereby undermining self-determination. Second, they can
carry an informational content which can be negative. For instance, in the model
by Bénabou and Tirole [2002a], promising a high bonus in case of success may
give a signal to the agent that the task in question will be difficult and that he is
unlikely to succeed. For as long as rewards are given, the agent is motivated to
make efforts. In the meantime, rewards lower his self-confidence. At the moment
that rewards are withdrawn, the self-confidence effect persists and induces less
motivation than before. Interestingly, there is a considerable body of evidence
showing exactly this pattern (see Deci and Ryan [1985]).
The hidden cost component of rewards have not been found in studies where
unexpected rewards are used (e.g., Deci, Koestner and Ryan [1999]). This is
predicted by the basic model of this chapter: in any equilibrium, the principal has
no incentive to give a discretionary bonus that decreases motivation in the next
period. Thus, the rewards bring good news. Besides, they cannot be controlling
since they are discretionary. However, with a slight modification, the model can
replicate a negative correlation between rewards and self-confidence. One key
condition is examined when a reward can lower self-confidence in equilibrium.
This is the case when effort and ability are substitutes. Now, the bonus is a
signal that the outcome was a failure. This reduces his self-confidence in his
ability, which he compensates by making more efforts. However, even though the
bonus reduces self-confidence, it stimulates the agent to work harder because he
realizes he must make efforts in order to succeed.
2 Unexpected
some probability.
in quotation signs, because in equilibrium the agent foresees the payment of a bonus with
Chapter 6. The Hidden Rewards of Rewards
132
The rest of the chapter is organized as follows. Section 6.2 presents the model
and discusses the main assumptions. The derivations of equilibria are given in
section 6.3. Section 6.4 describes some qualitative properties of the equilibrium.
The results are discussed in section 6.5. After that, the results are related to the
results of related literature in section 6.6. Finally, section 6.7 concludes.
6.2 The model
6.2.1 Preliminaries
This section describes the general setup of the game. Some of the basic ideas are
closely related to the work by Bénabou and Tirole [2002a] and Souvorov [2003].
Where assumptions differ, this is made clear in the text.
In the game, there are two players: a principal and an agent. There is a finite
number of periods, for simplicity set to two. The case with an infinite horizon has
been studied briefly by Bénabou and Tirole [2002a] in a slightly different model.
The agent has to decide on his effort level. He chooses to make efforts or not:
e ∈ {0, 1}. If the agent undertakes the task, i.e. e = 1, he incurs a cost of c in
terms of disutility. Depending both on effort, e, and ability, θ, the outcome of
the effort can be either a success, S, or a failure, F . The probability of success is
given by:
prob(S | e) = eθ.
(6.1)
In other words, ability and effort are complements. No effort induces a failure
with certainty. In case of success, it yields a payoff equal to V to the agent, and
W to the principal. A failure yields a payoff equal to zero for both. Both parties
are risk-neutral and the agent is protected by limited liability.
The principal has to select a reward policy. In each period, he can offer a bonus
b ∈ R+ to the agent.
6.2.2 The main assumptions
Most of the above description is relatively standard for a principal-agent game.
The model, however, departs from most conventional models in several respects.
Each of these are discussed in more detail.
Imperfect self-knowledge. Although it is not a usual assumption, the idea that
people have only imperfect knowledge about their personal characteristics is plau-
The model
133
sible (see for instance Bénabou and Tirole [2002b]). First because retrospective
evaluations of past utilities are known not always to be reliable (Kahneman
[1994]). Thus, based on retrospection, people make incorrect estimates about
how they will feel about certain matters. Moreover, some situations are new to
people. In this case, they do not have enough information about themselves to
infer their ability. Someone who tries to quit smoking for the first time is unlikely
to be able to guess how persistent he will be. This requires some learning, but
learning opportunities are usually limited.
Imperfect self-knowledge in the current context means that people are not
perfectly informed about their ability. They cannot foresee their ability to make
a success out of it. The task they have to undertake is for example relatively
new to them, or they have forgotten how well they did on this or a comparable
task in the past. They do form an estimate about their ability. Based on this
estimate, they form an estimate of their chances to succeed, which represents
their self-confidence.
To make things concrete: suppose that the agent can be either one of two
possible types, a high type with ability θH or a low type with ability θL < θ H.
His prior on being a high type is given by ρ. His self-confidence is then given by:
ρθH + (1 − ρ)θL .
(6.2)
Clearly, self-confidence is increasing in ρ. In the remainder of the chapter the
parameters θH and θL are kept fixed, and with slight abuse of terminology, selfconfidence is identified with the parameter ρ.
Non-contractibility of the bonus. The principal has the possibility to give a
reward b to the agent. However, a crucial assumption in the model will be that
the outcome is not observable to the agent or an outside party. The outcome is
therefore private information to the principal. It follows that a reward contingent
on the outcome cannot be specified in a contract, because the agent or third party
would not be able to verify the truthfulness of the principal’s claim. That is, the
principal can always report a failure and no party can contest this claim.
The non-contractibility is one of the main departures from the model of Bénabou and Tirole [2002a]. They have analyzed the case where a contract can be
written that specifies the bonus in advance. Of course, they also have to assume
that the output is verifiable to the agent. The case of noncontractibility is interesting because first, in reality there are many situations where the bonus is indeed
Chapter 6. The Hidden Rewards of Rewards
134
noncontractible, and second, evidence from experiments show that the relation
between bonuses and motivation differs depending on whether or not a bonus is
specified in advance.
Intrinsic motivation. Even though no contract that specifies a bonus can be
written, it is still assumed that agents have a motivation to make efforts. Economists usually takes rewards as the motivation to work. According to Frey [1997],
many psychologists emphasize that the motivation to undertake a task can come
from within the person. If they are motivated without apparent reward or environmental control, they are said to be intrinsically motivated. In the words of
Deci and Ryan [1985, 43]:
”Intrinsic motivation is the innate, natural propensity to engage one’s
interests and exercise one’s capacities, and in so doing, to seek and
conquer optimal challenges. Such motivation emerges spontaneously
from internal tendencies and can motivate behavior even without the
aid of extrinsic rewards or environmental controls”.
It is undisputed that people are intrinsically motivated to do certain things:
playing football, solving a puzzle, the list is endless. An assumption in this chapter
is that people are indeed motivated for the task they have to undertake, even if
they get no current rewards. This is not a completely innocent assumption. Even
if people are intrinsically motivated to perform certain tasks, it does not follow
that they are intrinsically motivated to do all possible thinkable tasks. However,
the assumption is not crucial in the sense that the agent may also be motivated
for expected rewards in the future, despite the absence of current rewards. The
model allows for both interpretations.
The motivation of the agent is modeled as the value V in the model. To make
things interesting, one additional assumption has to be made on V, namely that
it cannot be directly observed. In other words, it is assumed that the agent is
motivated to do a task for which the benefits come later in life. Thus, one can
interpret V as the discounted value of payoffs later in life, be it extrinsic or
intrinsic. The agent may be a pupil learning to play the piano. First, he needs to
practice all kind of chords, a rather dull activity. The reward only comes when
he is able to play a decent piece. The agent may also be a student studying for
an exam, or writing an essay, not pleasant tasks for many people. His benefits
The model
135
may be to get a job afterwards that he really likes. Or he may be a worker, who
undertakes the task with the prospect of getting a promotion afterwards.
Asymmetric information. As explained, the agent is not sure about his ability
to bring the task to a successful end. Moreover, the focus in on situations where
even afterwards he does not get to know directly for sure whether it was a success
or a failure. He only gets an imperfect signal about the outcome. On the other
hand, the principal is able to observe the outcome. For instance, the pupil learning
to play piano cannot really tell whether he is talented after a few sessions, but the
principal can tell, having seen many pupils trying before this pupil. The same is
true for the student, whose grade will only be imperfectly informative about his
ability. This is certainly the case where the grade is dependent on the subjectivity
of the teacher, as with an essay. For a worker, it may be the case that this is the
first time he undertakes the task, or that his task is only a small part of a bigger
whole he is part of, so that he is not able to judge the outcome based on his own
information only.
Note that this assumption is contrary to most conventional principal-agent
models, where the agent has more information rather than less. For example, in
the classic job-market signalling model of Spence [1973] it is the agent who knows
his ability, whereas the principal only knows the distribution of abilities among
the population.
The private signal that the agent gets is given by σ ∈ [0, 1]. This signal has a
conditional distribution G(σ | y) = Gy (σ) and density g(σ | y) = gy (σ), where y
is the outcome of the task: y ∈ {S, F }. A higher σ is interpreted as good news in
the sense that it is more likely that a success has occurred. To capture this idea,
it is assumed that the likelihood function l(σ) with
l(σ) ≡
gS (σ)
,
gF (σ)
(6.3)
is an increasing function in σ. This is the monotone likelihood ratio property
(MLRP).
The next section examines equilibrium behavior of the principal and agent. To
focus on interesting cases, the following additional assumptions are made.
Assumption 3 Were the agent to know his type, then he would only undertake
the task without a bonus if he is a high ability type: θ L V < c < θH V.
Chapter 6. The Hidden Rewards of Rewards
136
As will be demonstrated shortly, no bonus is offered by the principal in the second
period. If assumption 3 did not hold, then either the agent would never work in
period 2, or he would always work, independent of his self-confidence. In both
cases, there is no role for the principal to increase self-confidence. Thus, no bonus
would be given in the first period either.
Furthermore, the following restriction is put on the likelihood ratio:
Assumption 4 The likelihood ratio l(σ) is continuous in σ and has full support on [0, +∞). Furthermore, the monotone likelihood ratio property (MLRP) is
satisfied: l(σ) is everywhere increasing in σ.
The full support assumption simplifies matters. It is also used by Bénabou and
Tirole [2002a] and Souvorov [2003] in related settings. The MLRP is an essential
assumption in many models with asymmetric information.
Assumption 5 In period 1, the agent undertakes the task: e = 1 in period 1.
This last assumption is made to focus on the interesting aspect of the model,
which is the behavior of the agent in period 2. Although conditions can be derived
under which e = 1 is an equilibrium strategy in period 1, not much insight is
gained from doing that.
6.2.3 Timing and summary of the game
Each period is divided in two subperiods. In each first subperiod, the agent decides
to make effort or not: e ∈ {0, 1}. Effort costs c in terms of disutility. At the end of
the first subperiod, the principal observes the outcome (y ∈ {S, F }) and the agent
receives a private signal σ about the outcome. A success occurs with probability
eθ and gives a payoff V to the agent and W to the principal. In the second
subperiod, the principal determines his reward policy b ∈ R+ .
Note also the following: at the beginning of the game, both the principal and
the agent have the same prior ρ that the agent is of the high type, θ H . To simplify,
both of them observe effort. The signal σ is private information to the agent, but
the conditional distribution functions are common knowledge.
Equilibrium behavior
137
6.3 Equilibrium behavior
Consider then the behavior by the agent if he does not know his type. First note
that no bonus is ever offered to him in period 2. The intuition is simple: the bonus
is costly to the principal, and since it can have no impact on strategies played in
the past, he should never give a bonus at the final stage of the game. Therefore,
given his posterior on being a high type ρ0 , the agent works in period 2 if and
only if:
[ρ0 θH + (1 − ρ0 )θL ] V ≥ c.
(6.4)
His posterior on being a high type is depending on his prior of being a high type,
the private signal, and the bonus he received in the first period. If the agent did
not work in period 1, he receives no additional signals and his posterior remains at
his prior ρ. Suppose then that the agent did work. Suppose also, quite generally,
that the principal offers a reward bS with probability xS after success, and with
probability xF after a failure. Then, after a bonus bS and a signal σ, the agent
updates his prior on being a high ability type to ρS with:
ρS
ρ θH gS (σ)xS + (1 − θH )gF (σ)xF
=
.
1 − ρS
1 − ρ θL gS (σ)xS + (1 − θL )gF (σ)xF
(6.5)
Based on expression (6.4), the following proposition can be derived:
Proposition 13 Then there exist threshold levels of his initial self-confidence ρ̃S
and ρ̃F ≥ ρ̃S such that if the agent worked in the first period, was given a bonus
bS and observed a private signal σ, in the second period the agent: (i) never works
if ρ < ρ̃S , (ii) always works if ρ ≥ ρ̃F , and (iii) if ρ ∈ [ρ̃S , ρ̃F ), works if and only
if he has sufficiently good news (σ high enough).
Proof. All proofs are collected in the appendix.
In other words, if the agent has a sufficiently low self-confidence, he will not work
in period 2 whatever the signal or bonus he gets. If he has sufficiently high selfconfidence, he will work in period 2 even for the worst possible signal or bonus
he could get. For intermediate levels of initial self-confidence, he is sensitive to
the news he gets. If he gets good news, he will work in period 2. If he gets bad
news, he will not work in period 2.
All the specific thresholds are derived in the appendix. However, it is interesting
to note the following. Suppose that his initial self-confidence is in the intermediate
Chapter 6. The Hidden Rewards of Rewards
138
range: ρ ∈ [ρ̃S , ρ̃F ). He will work if and only if after observing bs , the signal σ
exceeds σ̃, with:
l(σ̃) =
xF
A(ρ).
xS
(6.6)
Here, A is a parameter depending on initial self-confidence3 . It is easy to show
that this parameter is decreasing in initial self-confidence. Thus, the threshold
signal σ̃ is decreasing in initial self-confidence (for a given principal’s policy).
This increases the set of private signals for which the agent will work in period 2.
In sum, for a higher initial self-confidence, it becomes more likely that the agent
will work in period 2. Note furthermore that the threshold signal is decreasing in
the probability that the bonus is paid after a success (xS ), which makes it more
likely that the outcome was a success, and increasing in the probability that the
bonus is paid after a failure (xF ), which makes it more likely that the outcome
was a failure.
Consider now the behavior by the principal in period 1. For ρ ∈
/ [ρ̃S , ρ̃F ), it
should be clear that the principal gives no bonus in period 1, since he is not able
to influence behavior in period 2. From here on, the focus will therefore be on
ρ ∈ [ρ̃S , ρ̃F ). For these values of ρ, the principal may try to signal through a bonus
that the agent was successful in period 1. The reason is that for this interval of
initial self-confidence, it can happen that the agents works and is successful, but
receives a private signal that is below the threshold σ̃. This possibility is less
likely if the principal increases the probability of giving a bonus bS after success,
or decreasing the probability of giving a bonus bS after failure.
However, there are typically many perfect Bayesian equilibria in this game. This
is a common feature of signalling games. Some of these equilibria are less reasonable than others, primarily because out-of-equilibrium beliefs are unrestricted. To
reduce the set of possible equilibria, it is useful to restrict the set of possible outof-equilibrium beliefs. Here, a relatively standard refinement is applied, which is
the Never a Weak Best Response (NWBR) refinement for signalling games, which
is, in the current context, equivalent to the universal divinity criterion (see Cho
and Kreps [1987] and Fudenberg and Tirole [1991]). A general definition is given
3 For
the values of ρ in the interval from ρ̃S to ρ̃F , A(ρ) is equal to the ratio of the expected loss from
working after a failure to the expected gain from working after success. In the absence of any intermediate
information the agent would work if and only if A(ρ) ≤ 1.
Equilibrium behavior
139
in the appendix. Reformulated for the current model, the refinement requires the
following:
Assumption 6 Fix an equilibrium outcome. For any out-of-equilibrium bonus b̂,
let σ̄ F (b̂) be the agent’s reaction4 to this bonus that makes the principal indifferent
between the expected payoff following b̂ (the agent playing σ̄F (b̂)) and his expected
equilibrium payoff if a failure has occurred. Then, if for this reaction σ̄ F (b̂) the
principal strictly prefers his expected equilibrium payoff to the expected payoff
from deviating to b̂ if a success has occurred, then the agent should believe that
a failure occurred after getting b̂. If for this reaction σ̄ F (b̂) the principal would
strictly prefer to deviate to b̂ if a success has occurred, the agent should believe
that a success has occurred after getting b̂.
The intuition behind the assumption is that if the principal wants to deviate to
an out-of-equilibrium bonus b̂ after success for any agent’s reaction making him
want to deviate after failure, the agent should believe that success is (infinitely)
more likely. Similarly for the failure. That is, given an equilibrium outcome, if
the principal expects σ̄ F (b̂) after deviating to an out-of-equilibrium bonus, and
he would strictly prefer to deviate after a success but be indifferent after a failure,
the agent should believe a success has occurred.
Under the assumptions made, the equilibrium is unique (see the appendix).
Depending on initial self-confidence, the equilibrium is either a pooling equilibrium or a semi-separating equilibrium. In the pooling equilibrium, the principal
always offers the same reward. In the semi-separating equilibrium, the principal
always offers the same reward after success, but randomizes between two rewards
after a failure. These two equilibria are examined in detail below.
Before continuing, the following lemma will prove to be helpful. Define σ̃ as the
threshold signal such that the agent works for all σ ≥ σ̃. Note that in equilibrium
this threshold signal depends on the bonus because the bonus provides additional
information. One can therefore write σ̃ = σ̃(b). Then:
Lemma 6 In any equilibrium, for b1 > b2 , it must be that σ̃(b1 ) < σ̃(b2 ).
In other words, a higher bonus increases the likelihood of effort. This is easy
to see. Suppose b1 and b2 are equilibrium bonuses but σ̃(b1 ) > σ̃(b2 ). Then a
loosely, we shall call σ̄(b) (or σ∗ (b)) the agent’s strategy to work after getting bonus b if and
only if his signal exceeds σ̄(b).
4 Somewhat
Chapter 6. The Hidden Rewards of Rewards
140
lower bonus increases the likelihood of effort. Clearly, this makes the principal
unambiguously better off so that b1 could not have been an equilibrium bonus.
6.3.1 A pooling equilibrium
Suppose that there is a pooling equilibrium, with xS = xF = 1. Thus, the same
bonus bonus b̃ = bs is always given independent of the outcome. Suppose also that
given this bonus, the agent only works for signals exceeding σ̃ > 0, where l(ρ̃) = A
is determined by (6.6). Denote by θ̂F and θ̂S the estimates by the principal of the
agent’s chances to succeed in period 2, conditional on failure and success in the
first period. Thus:
θ̂y = prob(S in period 2 | y in period 1).
(6.7)
It is assumed that the agent tried in period 1. The exact probabilities are given
in the appendix. The expected payoffs for the principal is then given by:
θ̂y (1 − Gy (σ̃))W − b̃.
(6.8)
Assume that the principal deviates from the equilibrium strategy, and offers a
bonus b̂ = b̃ + ε. Let σ̂ be the agent’s reaction to this offer which makes the
principal indifferent between deviating or not after failure:
θ̂F (1 − GF (σ̃))W − b̃ = θ̂F (1 − GF (σ̂))W − b̂,
(6.9)
ε = θ̂F (GF (σ̃) − GF (σ̂))W.
(6.10)
or
For instance, the left-hand side of equation (6.9) gives the expected profit of the
principal by paying b̃ knowing that a failure has occurred. The probability of
success in the second period is given by θ̂ F and the probability that the agent
works is given by 1 − GF (σ̃).
The question then is: can given this bonus b̂ and the corresponding σ̂, the bonus
b̃ be an equilibrium? This depends on the beliefs of the agents after observing b̂.
If the agent believes that a bonus b̂ is given after a success, the principal would
be able to achieve an increase in the probability of effort at arbitrarily small cost
ε. Obviously, the principal then has incentives to deviate to b̂, and b̃ cannot be an
equilibrium in this case. According to assumption 6 (NWBR), it must be that:
θ̂S (1 − GS (σ̃))W − b̃ ≥ θ̂ S (1 − GS (σ̂))W − b̂.
(6.11)
Equilibrium behavior
141
If this inequality did not hold, the agent should believe that a success has occurred
according to the NWBR assumption. This cannot be an equilibrium. In other
words, if the principal is indifferent between a bonus b̂ and b̃ after a failure,
he should not be better off with a bonus b̂ after a success. This is a necessary
condition.
Equations (6.9) and (6.11) combined yield:
θ̂ S (GS (σ̃) − GS (σ̂)) ≤ θ̂F (GF (σ̃) − GF (σ̂)).
(6.12)
Dividing both sides by σ̃ − σ̂ (note that σ̂ is necessarily smaller than σ̃ for a
positive ε) and taking the limit ε → +0 one gets θ̂S gS (σ̃) ≤ θ̂F gF (σ̃) or:
l(σ̃) ≤
θ̂ F
θ̂ S
.
(6.13)
Conversely, assume that (6.13) is satisfied and consider a possible deviation b̂ =
b̃ + ε. Since the MLRP implies that5 :
GS (σ̃) − GS (σ̂)
< l(σ̃),
GF (σ̃) − GF (σ̂)
(6.14)
for any σ̂ < σ̃, condition (6.13) together with assumption 6 are sufficient to insure
that the agent will believe in failure after receiving any out-of-equilibrium bonus
b̂ and so the principal has no incentive deviate to a higher bonus. In the appendix
we prove that the equilibrium bonus is zero so the principal cannot deviate to
a slightly lower bonus. In sum, for the proposed pooling equilibrium to exists, a
necessary and sufficient condition is (6.13).
For later reference, note that condition (6.13) and the condition that l(σ̃) = A
imply that A ≤ θ̂F /θ̂S .
6.3.2 A semi-separating equilibrium
Consider next a semi-separating equilibrium where the principal offers b̃S after
success and randomizes between b̃S and b̃F after failure. In this case, a bonus b̃F is
only given in case of failure and therefore perfectly reveals a failure. It follows that
5 To
prove this: suppose l(x) < l(z) ∀ z ∈ [x, y]. Then, since l(z) ≡ gS (z)/gF (z), gF (z)l(x) < gS (z).
R
R
This implies xy gF (z)l(x)dz < xy gS (z)dz. Integrating out yields: l(x) < [GS (y) − GS (x)] / [GF (y) − GF (x)] .
Similarly, for l(z) < l(y) ∀ z ∈ [x, y], it follows that [GS (y) − GS (x)] / [GF (y) − GF (x)] < l(y). Note in particular
that for y > 0, GS (y)/GF (y) < l(y).
Chapter 6. The Hidden Rewards of Rewards
142
b̃F = 0, since there is no reason for the principal to incur a cost when conveying
a negative signal.
Next note that in equilibrium the principal must be indifferent between b̃S and
b̃F after a failure, otherwise he would not be willing to mix. In other words:
θ̂F (1 − GF (σ̃S ))W − b̃S = 0,
(6.15)
where σ̃S is the threshold signal for which an agent works after a bonus b̃S . Note
that after b̃F = 0 the agent does not work (recall that θL V < C) so the payoff for
the principal is zero. This condition determines the bonus b̃S .
Moreover, the principal should not want to deviate to a bonus slightly above
or below b̃S . Following similar logic as above, the principal does not want this in
case the agent would believe a failure occurred after observing the deviation from
the equilibrium bonus. Suppose first that the principal deviates to b̂S = b̃S − ε.
Assumption 6 implies that if
θ̂F (1 − GF (σ̂S ))W − b̂s = 0,
(6.16)
θ̂S (1 − GS (σ̂ S ))W − b̂S ≤ θ̂S (1 − Gs (σ̃S ))W − b̃S
(6.17)
it must be that
Again, if this inequality did not hold, the agent should believe that a success has
occurred by the NWBR assumption. This cannot be an equilibrium.
Together, these conditions imply:
l(σ̃S ) ≥
θ̂F
θ̂ S
.
(6.18)
(Recall that a deviation to a smaller bonus is considered, hence σ̂ > σ̃).
Now suppose σ̃ S = 0. The above condition implies l(0) = 0 ≥ θ̂F /θ̂ S . However,
θ̂F /θ̂S > 0 so this case can be ruled out. Then consider σ̃ S > 0, so that the agent
does not always work after a bonus b̃S . Then, the principal should also not be
willing to deviate to a higher bonus b̂S = b̃S + ε to separate the success outcome.
Like in the case of pooling, this implies:
l(σ̃S ) ≤
Hence, combining (6.18) and (6.19) gives:
θ̂F
θ̂ S
.
(6.19)
Rewards, self-confidence, and motivation
143
l(σ̃S ) =
θ̂F
θ̂S
(6.20)
as the only possibility. The agent’s reaction is given by
l(σ̃S ) = x̃F A.
(6.21)
Condition (6.20) determines σ̃S , (6.21) defines x̃F and (6.15) determines b̃S . Finally note that conditions (6.20) and (6.21) imply that θ̂F /θ̂S ≤ A since x̃F ≤ 1.
6.4 Rewards, self-confidence, and motivation
The following is the main proposition of the chapter, which relates the reward to
initial self-confidence and motivation in period 26 :
Proposition 14 Under assumptions (3)-(6), there always exists is a unique continuation equilibrium depending on the initial self-confidence. In particular, there
exists a threshold level ρ∗ such that for values ρ ∈ [ρ̃S , ρ̃F ):
(i) for ρ < ρ∗ , the unique equilibrium is a semi-separating equilibrium in which
the principal always offers a bonus bS = θ̂ F (1 − GF (σ̃S ))W after success
( xS = 1), and randomizes between bS and bF = 0 after a failure with
l(σ̃)
and 1 − xF respectively. The threshold σ̃S is positive
probabilities xF = A(ρ)
and determined by l(σ̃ S ) =
θ̂F
,
θ̂S
and σ̃ F = 1.
(ii) for ρ ≥ ρ∗ , the unique equilibrium is a pooling equilibrium where no bonus
is ever offered. The threshold σ̃ is positive and determined by l(σ̃) = A(ρ).
The threshold level ρ∗ is determined by the ρ for which7
θ̂F (ρ∗ )/θ̂S (ρ∗ ) = A(ρ∗ ).
(6.22)
6 The proposition states the equilibrium conditions for values of ρ such that ρ ∈ [ρ̃ , ρ̃ ). Recall that it was
S F
already established that for values of ρ ∈
/ [ρ̃S , ρ̃F ) no bonus is ever offered. For ρ < ρ̃S , σ̃S = σ̃F = 1 (the agent
never works) and for ρ ≥ ρ̃F , σ̃S = σ̃F = 0 (the agent always works).
7 The existence of a point ρ∗ is obvious: A(ρ) decreases from infinity to 0 on the interval [ρ̃ , ρ̃ ] and
S F
θ̂ (ρ)
r(ρ) = F
is positive and bounded away from 0 on this interval. Implicitly we assume uniqueness as well. For
θ̂ S (ρ)
this it is sufficient (but not necessary) to require that r(ρ) be non-decreasing on the relevant interval. Multiple
ρ∗ s would slightly modify the proposition in a straightforward manner.
Chapter 6. The Hidden Rewards of Rewards
144
That the unique equilibrium for sufficiently high initial self-confidence is a pooling
equilibrium with no bonus is intuitive: if self-confidence is high, the agent is likely
to work in the second period, and it becomes too costly for the principal to signal
a success. The main point of the proposition is however that there is a region
where the principal does have an incentive to give a bonus, and that this bonus
increases self-confidence. In this region, the agent is relatively unlikely to make
efforts in the second period. In this case, the principal has an incentive to make
a costly signal to the agent to make clear to him that a success has occurred.
It is also possible to show that for ρ < ρ∗ the probability of a reward increases in
initial self-confidence. Since θ̂F /θ̂S is increasing in ρ, so must l(σ̃S ). It is then easily
seen that xF must be increasing, since A(ρ) is decreasing in ρ. The probability
that an agent works, on the other hand, is decreasing, as σ̃ S increases in ρ. The
change in the size of the bonus for a higher initial self-confidence is ambiguous.
There are two opposing effects: first θ̂F increases, since a higher probability of
a high type increases the success of the agent in the second period. Secondly,
1 − GF (σ̃ S ) decreases since σ̃S increases. The total effect depends on the relative
sizes of these two opposing effects.
The size of the bonus in the region where self-confidence is relatively low, that
is ρ < ρ∗ , is proportional to the payoff for the principal in case of success, W.
This means that the scope of applications is not limited to situations where the
stakes are high for the principal. For example, it would be enough if the principal
derives a small benefit from observing a successful performance of the agent, say
out of altruistic feelings. For smaller stakes, the corresponding equilibrium bonus
will be lower.
6.5 Discussion
When contracts are absent in a relationship, one easily ends up with the argument
that no bonus will ever be given, and neither that efforts will be made. The cause
is the strong backward induction argument: the agent knows that the principal
has no incentive to give a reward in the last period and so he makes no effort,
after which the principal realizes that rewarding in the before-last period makes
no sense, and so the agent will not work in that period either, and so on until the
very first period. This chapter sheds some light on why rewards and efforts may
be observed after all.
Discussion
145
There is no shortage of empirical and experimental evidence that shows the
existence of rewards which are not conditioned on performance, and also that
there is a positive relationship between rewards and efforts. Akerlof [1982], for instance, has noted that labour markets can often be characterized as gift exchange
relationships. The employers give wages above the minimum wage, and workers make more efforts than is required. Laboratory experiments show the same
positive relationship between wages and efforts, even in the absence of explicit
performance incentives (Fehr and Gächter [2000]). Deci and Ryan [1999] survey
the psychological literature on rewards and intrinsic motivation. They present
some studies which find a positive effect, although not all studies which used
unexpected rewards find a positive effect, and on average they find no significant
relation8 .
The positive relationship between rewards and efforts is also called positive
reciprocity. Chapters 2 and 3 provide a more detailed exposition and extend
the phenomenon to other environments. These chapters give other explanations
of this relationship. For example, it is advanced that people are reciprocal by
virtue of their fair nature: they are driven by the moral obligation to reward
generous behavior by generous behavior (see e.g. Falk et al. [1999] and section
2.3.3). Another possibility is that people care about social approval and that
generous behavior elicits generous behavior (see chapter 3). The current chapter
adds another explanation to the existing literature, by focusing on the role of
self-confidence.
Obviously, the proposed mechanism can only be valid as long as the main
assumptions are satisfied. An important assumption is that the principal has more
information about the expected payoffs than the agent. This makes the theory
more applicable to situations where agents are in their learning phase: at school
or at new jobs. A second important assumption is the sorting condition that is
implicit in the model. The principal must obtain a higher marginal benefit from
rewarding an agent after a success than after a failure. Otherwise, the principal
would be tempted to reward the agent after a failure as well, disturbing the
proposed equilibrium.
8 However,
in these studies it is not clear to the participants what the benefits of the experimentator are.
The setup of the experiments do therefore not completely fit the current model.
Chapter 6. The Hidden Rewards of Rewards
146
The model is also extendable to other situations with asymmetric information.
For example, it extends to situations where the agent is unsure about his own
payoff rather than his ability. Another possibility is that the agent cares about
the principal’s payoff (e.g. through altruism), but is unaware of how much utility
the principal derives from his effort.
6.6 The hidden costs of rewards
So far the focus has been on how noncontracted rewards stimulate motivation. By
contrast, the papers by Bénabou and Tirole [2002a] and Souvorov [2003] study
how rewards can decrease motivation. As argued earlier, the main difference with
their approach is that they consider bonuses that are specified in a contract.
To sketch the argument: if a principal observes that the agent has low ability,
he also expects him to have low self-confidence. He therefore proposes a high
powered contract which specifies a high reward contingent on success to motivate
the agent anyway. This makes the agent realize that he must be of low ability,
which lowers his self-confidence. Whenever rewards are withdrawn, the agent will
be less motivated.
The mentioned papers are initiated because there is much evidence that this
effect occurs. For example, in one experiment it is found that children who were
paid for engaging in an activity, showed less interest in the activity once rewards
were withdrawn than children who were never rewarded for the activity (Lepper
et al. [1973])9 . Deci and Ryan [1999] survey the literature and find that such
crowding-out of motivation also comes out of a meta-analysis of more than one
hundred earlier studies. In sum, there is a rich body of experiments showing that
there are hidden costs of rewards10 .
How much motivation is crowded out depends to a great extent on the nature
of the reward (Deci and Ryan [1985]). For example, rewards contingent on performance have an effect on motivation, but the effect of rewards contingent on
fulfilling the task are less profound. No such an effect has been found for experiments where rewards were unexpected, although the evidence is mixed and the
results are small and on average insignificant.
9 Interestingly,
10 See
when the reward was unexpected, there was a slight (yet insignificant) increase in interest.
also Kohn [1993], Deci and Ryan [1985], Frey [1997], Frey and Jegen [2002], and chapters 2, 3 and 7.
147
Conclusions
It is possible to replicate the negative correlation between rewards and selfconfidence in the current model. In the model of section 6.2, the principal has more
to gain from rewarding an agent after a success, and only then it is worth making
a costly signal. Thus, the reward is good news for the agent and raises his selfconfidence. However, so far it is assumed that efforts and ability are complements,
so that the principal indeed wants to increase the agent’s self-confidence. There
are, however, also cases where a higher self-confidence would be a bad thing from
the principal’s viewpoint, and the results of the model would be reversed: the
principal would have an incentive to offer a reward to signal a failure, hence
reducing self-confidence.
Consider for example the case where the agent has to perform a task that
has again only two possible outcomes: a failure or a success. Now assume that
the ability and effort are substitutes rather than complements. For instance, a
student may be intelligent enough to pass an exam without any efforts. Less gifted
students can compensate their lack of ability by making more efforts and study
hard. In any case, the parent would rather make the student work hard to avoid
any risks of failure. By giving a reward he could say: ”Look, here’s a reward to
show that you have failed, you’d better work hard next period to pass.” Although
this lowers his self-confidence, it would increase his motivation. Hence, no hidden
costs of unexpected rewards should be expected11 .
6.7 Conclusions
Studies by psychologists show that rewards can undermine motivation. This has
stimulated economists to examine the effects of rewards in more detail. One in
particular interesting contribution is the paper by Bénabou and Tirole [2002a]
where the focus is on the role of self-confidence. This chapter also examines the
effect of rewards on self-confidence. The focus is on rewards that are not specified
in contracts. This answers several questions, such as why discretionary rewards
are used and how they can stimulate motivation.
1 1 Consider
for instance the following version of the model. The agent’s payoff is given by λV − ec, with
λ = 1 if and only if e + θi ≥ ϕ. Thus, λ = 1 is a success and needs either effort or a high ability. Assume that
ϕ < 1, and V > c so that effort is enough to pass. However, an able agent can pass without effort whereas a low
ability agent cannot: θH > λ > θL . If no effort is made in the first period, and the outcome was not a success,
the principal may want to signal through a bonus that the agent has low ability and should make efforts in the
second period.
Chapter 6. The Hidden Rewards of Rewards
148
Rewards can motivate agents by signalling a success, increasing self-confidence.
Rewards can also decrease self-confidence in the special case where efforts and
ability are substitutes. In accordance with the empirical and experimental evidence, no negative relation between rewards and motivation is found for such
unexpected rewards.
This chapter is only one of the first few attempts to formally study the interaction between rewards and self-confidence. Future work should generalize some
of the assumptions. For example, the role of shirking is central in many principal agent models. Here, it is simply assumed that effort is observable. Another
straightforward extension is to include more than two periods. This would shed
light on the dynamics of rewards and self-confidence in the spirit of Souvorov
[2003]. For instance, in a richer framework, the agent may benefit from pretending his confidence is low in order to get a bonus. In addition, people may come
to expect another reward once they received one in a previous period. This may
complicate the analysis somewhat as this changes their reference point. Furthermore, it would be interesting to examine the relation between promised bonuses
(specified in contracts) and discretionary bonuses. Promised bonuses can undermine motivation but at least specify when a bonus will be given. Discretionary
bonuses increase motivation to work after a bonus, but insofar as they are unexpected they cannot motivate agents in the first period. Thus, perhaps there is
something such as an optimal mix of different kinds of rewards, consisting both
of announced as well as unexpected bonuses.
Appendix
149
6.8 Appendix
Proof of proposition 13.
With the strategy of the principal is to give bS after success with probability xS
and after failure with probability xF , the agent updates his prior ρ to ρS after a
bonus bS . The expressions for ρS is given in the text (equation (6.5)). Consider
an agent who observes bS . He works if and only if:
[ρS θH + (1 − ρS )θL ] V ≥ c.
(6.23)
ρS
c/V − θL
≥
≡ φ.
1 − ρS
θH − c/V
(6.24)
This can be written as:
Define ρ0 to be ρS such that (6.24) holds with equality. Then, using (6.5), one
gets:
xF
gS (σ)
xF φ(1 − ρ0 )(1 − θL ) − ρ0 (1 − θH )
=
≡
A(ρ0 ).
(6.25)
0
0
gF (σ)
xS
ρ θH − φ(1 − ρ )θ L
xS
The denominator of the RHS is zero for:
φθ L
.
φθL + θH
(6.26)
φ(1 − θL )
.
φ(1 − θL ) + (1 − θH )
(6.27)
ρ0 = ρ̃S ≡
The numerator of the RHS is zero for:
ρ0 = ρ̃F ≡
The sign of the derivative of the RHS of (6.25) is equal to the sign of (θL −θH )φ <
0. Furthermore, it is easy to show that ρ̃S < ρ̃F . Finally, note that:
lim A(ρ) → −∞,
(6.28)
lim A(ρ) → +∞
(6.29)
ρ↑ρ̃S
and
ρ↓ρ̃S
In sum, for ρ < ρ̃S and ρ ≥ ρ̃F the RHS of (6.25) is negative and decreasing, and for ρ̃S < ρ < ρ̃F , the RHS of (6.25) is positive and decreasing. Since
gS (σ)/gF (σ) ≥ 0, there is no signal for ρ ∈
/ [ρ̃S , ρ̃F ) such that (6.25) holds with
equality. For ρ < ρ̃S the agent never finds it profitable to work. For ρ > ρ̃F , the
Chapter 6. The Hidden Rewards of Rewards
150
agent always finds it profitable to work. For ρ ∈ [ρ̃S , ρ̃F ) the agent works for any
signal σ > σ̃ with:
gS (σ̃)
xF
A(ρ).
(6.30)
=
gF (σ̃)
xS
Since it is assumed that gS (σ)/gF (σ) has full support on [0, +∞), the threshold
signal σ̃ always exists. Note that thresholds ρ̃S and ρ̃F do not depend on the
bonus bS .¥
Proof of proposition 14.
In proving the proposition, we first introduce some intermediate results in the
form of lemma’s and a corollary. At the end we give the precise definition of the
NWBR assumption, because this requires some additional notation of it’s own.
Lemma 7 If bonus b̃ is given in equilibrium with probability x̃S > 0 after success
and with x̃F > 0 after failure, and σ̃ is the agent’s reaction to the bonus (i.e. the
agent works if and only if he received a signal above σ̃), then l(σ̃) = x̃x̃FS A(ρ) and
either
• b̃ > 0 and l(σ̃) = r(ρ) or
• b̃ = 0 and l(σ̃) ≤ r(ρ),
with r(ρ) ≡ θ̂F /θ̂ S .
Proof. l(σ̃) = x̃x̃FS A(ρ) determines the agent’s optimal reaction to the principal’s
policy unless the agent would find it worthwhile to always work when offered b̃,
i.e. σ̃ = 0. In the latter case we should have l(0) ≥ x̃x̃FS A(ρ) in contradiction with
Assumption 4 which states l(0) = 0.
When b̃ > 0, for the principal not to be able (and a fortiori willing) to signal
that the agent has succeeded by deviating to b̃±ε for a small ε > 0, it must be the
case that l(σ̃) = r(ρ) (see the analysis of the pooling equilibrium in section 6.3.1).
When b̃ = 0, only deviations to b̃ + ε are relevant so the requirement reduces to
l(σ̃) ≤ r(ρ).
Lemma 8 In equilibrium only one bonus is offered after success.
Proof. Assume that b1 and b2 > b1 are offered after success with positive probability, and σ̃ 1 and σ̃ 2 are the corresponding agent’s reactions (σ̃ 1 > σ̃2 by Lemma
Appendix
151
6). The smaller bonus, b1 , must be offered after failure with a positive probability (otherwise the agent would always work after b1 and the principal would
never give the larger one, b2 ). For the principal not to be willing to separate the
successful outcome by offering b1 + ε, it must be that l(σ̃1 ) ≤ r(ρ). Then,
b2 − b1 = θ̂S (GS (σ̃1 ) − GS (σ̃2 )) < θ̂F (GF (σ̃1 ) − GF (σ̃2 )).
(6.31)
The equality in (6.31) comes from the principal’s indifference between b1 and b2 ,
and the inequality follows from l(σ̃1 ) ≤ r(ρ) and MLRP and implies that that the
principal strictly prefers to give b2 rather than b1 after a failure — a contradiction.
Corollary 2 At most two different bonuses are offered with positive probability
in equilibrium. There are three potential types of equilibrium:
A. pooling — the same bonus offered to both types;
B. semi-separating — the principal always gives b̃S after success and randomizes
between b̃S and b̃F 6= b̃S after failure;
C. separating — the principal always gives b̃S after success and b̃F 6= b̃S after
failure.
Each possible type of equilibrium is considered below.
A. Pooling equilibria
Some bonus b̃ is always given to the agent (xF (b̃) = xS (b̃) = 1), who works when
observes a signal σ above σ̃ = σ̃(b̃).
Case A.1 : b̃ = 0 and σ̃ = 0 (the agent always works).
For this equilibrium to occur it must be the case that
l(0) ≥ A(ρ).
(6.32)
By the full support assumption for the likelihood ratio, this is equivalent to ρ ≥
ρ̃F .
Case A.2 : b̃ = 0 and σ̃ > 0 (the agent does not always work).
Chapter 6. The Hidden Rewards of Rewards
152
See section 6.3.1 in the main text. Note that:
ρ(1 − θH )θH + (1 − ρ)(1 − θL )θL
,
ρ(1 − θH ) + (1 − ρ)(1 − θL )
ρθ2H + (1 − ρ)θ2L
=
ρθH + (1 − ρ)θL
θ̂F =
(6.33)
θ̂S
(6.34)
Case A.3 : b̃ > 0 and σ̃ > 0 (the agent does not always work).
As in case A.2, the principal should not want to deviate in case of success. Now
the condition for the principal’s inability to separate a success is more stringent:
l(σ̃) =
θ̂ F
θ̂S
.
(6.35)
Indeed, by the same reasoning as in the previous case it easy to show that
l(σ̃) ≤ θ̂ F /θ̂S is necessary and sufficient for the principal’s inability to separate
the success outcome by increasing b̃ by ε. According to similar logic, l(σ̃) ≥ θ̂F /θ̂S
is a necessary and sufficient condition for the principal not being able to signal
success through reducing the equilibrium bonus by ε. This leaves (6.35) as the
only possible case.
Besides (6.35), there is another condition for this equilibrium: the principal
should not wish to separate the failure outcome. In the previous case it was
satisfied trivially because increasing a bonus in order to decrease the probability
of effort is clearly not a good idea. Now, when the principal can reduce her
equilibrium bonus, this requirement may be restrictive. When (6.35) is satisfied,
any deviation will be interpreted by the agent a signal of failure. Hence, the
optimal deviation is to pay no bonus at all. This deviation will not to be profitable
if and only if
θ̂F (1 − GF (σ̃))W − b̃ ≥ 0
because the agent will not work if convinced in failure (remember that ρ < ρ̃F ).
Thus, in this non-generic case (i.e. when l(σ̃) = θ̂F /θ̂S for σ̃ determined from
l(σ̃) = A) there is a continuum of pooling equilibria with b̃ ∈ [0, θ̂F (1 − GF (σ̃))W ]
satisfying the NWBR criterion.
Case A.4 : b̃ > 0 and σ̃ = 0 (all agents work).
Appendix
153
Clearly, this case requires l(0) ≥ A(ρ). By the full support assumption for the
likelihood ratio, this is equivalent to ρ ≥ ρ̃F . It also requires l(0) ≥ θ̂F /θ̂S , which
can be proved by a a line of reasoning similar to the other cases above.
Semi-separating equilibria.
Case B.1 : Failure outcome semi-separated. See section 6.3.2 in the main text.
Case B.2 : Success outcome semi-separated.
This equilibrium is impossible: Assume that the principal pays b̃F after failure and
randomizes between b̃S and b̃F after success; the agent then always works after
b̃S and works when σ ≥ σ̃F after b̃F . Clearly b̃S > b̃F — otherwise the principal
would always pay b̃S , and σ̃F > 0 — otherwise the principal would always pay
b̃F . For the principal not to be able to separate a successful outcome by a bonus
b̃F + ε for arbitrarily small ε it must be that
l(σ̃ F ) ≤
θ̂ F
θ̂S
(6.36)
In case b̃F > 0, this should hold with equality since the principal should also not be
able to separate the success outcome by decreasing the bonus by ε. Furthermore,
the principal should be indifferent between b̃S and b̃F after a success otherwise
he should not be willing to randomize:
θ̂S (1 − GS (σ̃F ))W − b̃F = θ̂S W − b̃S ,
(6.37)
But (6.36) implies that if the principal is indifferent between giving b̃S or b̃F after
success, she should prefer to give b̃S after failure — a contradiction. Indeed, the
MLRP property of l(σ) implies that
GS (σ̃)
< l(σ̃).
GF (σ̃)
(6.38)
But then if the principal is indifferent between b̃S and b̃F after success, he prefers
b̃S after a failure:
θ̂F (1 − GF (σ̃ F ))W − b̃F < θ̂S W − b̃S .
Hence b̃F cannot be an equilibrium bonus.
Case B.3 : Both outcomes semi-separated
(6.39)
Chapter 6. The Hidden Rewards of Rewards
154
This equilibrium cannot occur. It implies that the principal must be indifferent
between b̃S and b̃F both after a failure and after a success. It is easy to show that
this is not compatible.
Separating equilibria.
In a separating equilibrium the principal gives b̃F = 0 after a failure (again,
there is no sense to incur any cost to send a negative signal) and b̃S > 0 after
success. The agent always works after b̃S and never works after b̃F . For this pair
of bonuses to be an equilibrium, the principal should not strictly prefer to give
b̃S after failure:
θ̂F W − b̃S ≤ 0.
(6.40)
If (6.40) were a strict inequality, then the principal could reduce b̃S by a small ε so
that (6.40) would still be satisfied and according to he NWBR criterion the agent
should believe success has occurred: the set of the agent’s reactions that make
the principal indifferent between giving 0 and b̃S − ε after failure — the empty set
— is strictly included in the set of reactions that make her indifferent between b̃S
and b̃S − ε after success. Hence, b̃S is uniquely determined:
b̃S = θ̂F W.
(6.41)
It must also be the case that the principal cannot separate the success outcome
by a bonus slightly lower than b̃S . A now familiar condition for this is
l(0) ≥
θ̂F
θ̂S
.
(6.42)
To see this, assume that the agent’s reaction σ̂ to an out-of-equilibrium bonus b̂
is such that the principal is indifferent between deviating to b̂ after a failure or
not:
θ̂F (1 − GF (σ̂))W − b̂ = 0.
(6.43)
Then we need the principal not to be willing to deviate after success:
θ̂S (1 − GS (σ̂))W − b̂ < θ̂S W − b̃S ,
(6.44)
θ̂S GS (σ̂) > θ̂F GF (σ̂).
(6.45)
or, using b̃S = θ̂F W
For (6.45) to be satisfied for all σ̂, a necessary and sufficient condition is (6.42).
Appendix
155
In sum, all equilibria other than those of proposition 14 require either that l(0) ≥
θ̂F /θ̂S or l(0) ≥ A(ρ) or both. Since θ̂F /θ̂S and A(ρ) are both positive for ρ ∈
[ρ̃S , ρ̃F ) and since l(0) = 0, these can be ruled out. The only exception is case A.3
which exists only for the point where θ̂F /θ̂ S = A(ρ) which has measure zero and
is therefore for simplicity ignored. ¥
Never a Weak Best Response in signalling games.
The following is a condensed presentation of the NWBR criterion. See Fudenberg
and Tirole [1991] for more details.
Suppose the principal chooses action a1 and the agent a2 . This notation facilitates comparison with the literature. We say that a principal is of type t if he
has observed outcome t. Then, fix an equilibrium, an out-of-equilibrium action
a1 , and let u∗1 (t) be the principal’s expected payoff in the proposed equilibrium if
he is of type t. Then define the set D(t, T, a1 ) to be the set of mixed-strategy best
responses α2 to action a1 and beliefs concentrated on T (a subset of the set of all
possible types Θ) that make type t strictly prefer a1 to his equilibrium strategy:
D(t, T, a1 ) = ∪µ:µ(T |a1 )=1 {α2 ∈ MBR(µ, a1 ) s.t. u∗1 (t) < u1 (a1 , α2 , t)},
(6.46)
for beliefs µ over different types.The set Do (t, T, a1 ) is defined as the set of mixedstrategy best responses that make type t indifferent:
D o (t, T, a1 ) = ∪µ:µ(T |a1 )=1 {α2 ∈ MBR(µ, a1 ) s.t. u∗1 (t) = u1 (a1 , α2 , t)}.
(6.47)
Then, a type-action pair (t, a1 ) can be deleted under the NWBR criterion if:
D o (t, Θ, a1 ) ⊂ ∪t0 6=t D(t0 , Θ, a1 ).
(6.48)
Type-action pair (t, a1 ) can be pruned if the (sequential) equilibrium response
to the out-of-equilibrium action a1 that makes t indifferent between his expected
equilibrium outcome and a deviation to a1 , makes some other type strictly prefer
to deviate to the out-of-equilibrium action. In other words, the agent is assumed
to believe that it is infinitely more likely that a1 has come from some other type
t0 .
7
Optimal Subsidies with Rationalizing Agents:
Subsidize Enough but Don’t Subsidize Too
Much
”He suddenly recalled how he had once in the past been asked, ”Why
do you hate so and so, so much?” And he had answered them, with
his shameless impudence, ”I’ll tell you. He has done me no harm. But
I played him a dirty trick, and ever since I have hated him.” F. M.
Dostoyevsky, The Brothers Karamazov, 1879.
7.1 Introduction
The typical economic approach is to take preferences as given and then study
how certain incentives can alter behavior in a desired way. Psychologists on the
other hand, have paid considerable attention to the formation of preferences,
trying to demonstrate that they are not stable. Preferences are not stable indeed,
sometimes even capricious (see the next section). According to psychologists, the
formation of preferences is nevertheless in many cases quite predictable.
While changing behavior by appropriate incentives has been subject to extensive investigation within the field of economics, changing preferences (or attitudes,
or tastes) has been left largely unexplored. In this chapter both effects are taken
into account. I wish to argue that a subsidy on environmental friendly goods
0I
am indebted to Riccardo Calcagno, Theo van de Klundert, Eloic Peyrache, Karim Sadrieh, Sjak Smulders,
participants at the European Economic Association (Lausanne, 2001), the ENTER Jamboree (Toulouse, 2002)
and seminar participants at Tilburg for very helpful comments.
Chapter 7. Optimal Subsidies with Rationalizing Agents
158
(green products) not only influences behavior but also attitudes towards such
goods. Although subsidies have a straightforward effect on buying behavior — a
higher subsidy giving more incentives to buy that good — the relationship between
subsidies and attitudes is surprising. The main result is that a low subsidy stimulates a positive attitude change towards the subsidized good. Yet, a high subsidy
does nothing to the attitudes of people. This result fits the experimental evidence
in the psychological literature well, reporting a negative relationship between rewards and attitude change (Aronson [1988]). It follows that after the removal of
a high subsidy behavior and attitudes are as in the status quo. But a low subsidy
induces attitude changes that can persist even after withdrawing the subsidy. I
therefore conclude that high subsidies are ’too much of a good thing’: they affect
current behavior but fail to affect attitudes and therefore future behavior.
The leading example of this chapter is the consumption of environmental damaging goods, but the model applies to other instances of goods with social externalities as well, such as production methods that involve child labour or cruelty
to animals (see also Chapter 5 for more on goods with social externalities). Typical of environmentally damaging products (as well as the other examples) is that
the externalities tied to these goods are considered to be socially or morally undesirable. Consuming these goods therefore poses a threat to our self-concept of
being decent people. Psychologists argue that in such a case consumers experience an unpleasant feeling. To reduce this unpleasant feeling, they will either try
to change their behavior, or — by rationalizing their choice — their attitudes1 .
That people can and do change their behavior is undisputed. But, as said, they
can change their attitudes as well. There are several ways to accomplish attitude
changes (see Baumeister [1998]). For example, Frey [1997] argues that when the
costs of following principles of environmental ethics are high, people find a lot of
reasons why they should desist from doing so. They can for instance highlight
the argument that their effect of buying environmentally damaging goods is negligible on the environment anyway. The mere fact that you contribute marginally
to a bad environment does not make you an indecent person, does it? Or you
may recall that article in the newspaper saying that the damaging effects on the
environment are staggeringly exaggerated. And if that is not enough, you can
1 In psychological terminology: the bad feeling is caused by cognitive dissonance — a discrepancy between
simultaneously held cognitions (beliefs, attitudes, opinions). Originated by Festinger [1957] the theory has been
found to be mostly relevant in situations where one’s self-concept is violated by one’s behavior (Aronson [1988]).
159
Introduction
always buy yourself the book ”The Skeptical Environmentalist” by Bjørn Lomborg, which plainly denies any signs of an impoverishing environment. Just try to
ignore those other articles claiming the opposite. Possibilities abound. It should
therefore come at no surprise that self-serving biased information processing is
indeed systematically found (see the next section).
The benefits of changing attitudes towards environmental damaging products
is to take away the associated unpleasant feeling. This goes at some psychological
and/or psychic costs: information has to be gathered and mental efforts have to
be put into rationalizing choice. It is precisely the impact of subsidies on the costbenefit structure that drives the main result of this chapter. In words (formalized
in the next section): both a high as well as a low subsidy on green products
might induce a change in behavior towards buying these goods. But it does not
in both cases pay off to change your attitude. A high subsidy provides enough
justification of itself to buy green products. Even if you think that it does not
help the environment, it is still attractive to buy these goods due to their low
price. A low subsidy, on the other hand, does not give such a justification of
itself. Why did you until recently refuse to buy green products (perhaps claiming
that buying them has no significant merit) but consume them now that they are
slightly subsidized? That does not make sense. But it would start making sense
if you concentrate on the argument that, though each individual as such does
not contribute anything noteworthy, for the society as a whole it would make
a huge difference if everybody would behave the same way. Be supportive of
Kantian principles of morality. And, think of the social approval you will get for
the noble act of buying green products (see Chapter 3). Accordingly, the people
who received a relatively small reward are likely to revise their attitude most.
The rest of the chapter is organized as follows. In the next section I discuss
in more detail the basic premises of the model. Some evidence for the model is
provided in section 7.3. The formal model is presented in section 7.4, together
with the results. In the subsequent section I briefly discuss the robustness of
the results. Finally, section 7.5 discusses the results, relates them to the existing
literature, and concludes.
Chapter 7. Optimal Subsidies with Rationalizing Agents
160
7.2 Preference management
When a person deliberately harms someone else, he experiences an unpleasant
feeling: sentiments of guilt. This unpleasant feeling stems from a discrepancy
between held beliefs (”I am a decent person”) and actual behavior (”I harmed
her”). In the psychological literature this feeling has been labeled dissonance
(see also footnote 1). For ease of comparison with the existing literature, in the
remainder I stick to this terminology.
It is not so controversial to think that people experience dissonance when their
behavior imposes negative externalities on society. More controversial is that, to
reduce this unpleasant feeling, people can manipulate their preferences in a way
that serves their interest best. In other words, that people would ’rationalize’
choice. It is one of the most basic assumptions of theories of rational choice that
preferences are stable. Theories of habit formation depart from this by assuming
instead that future tastes are dependent on current consumption, for example as a
consequence of addiction or learning. These theories therefore allow for changing
tastes over time, but typically in a slow manner and in the direction of getting
to appreciate what you consumed in the past. Hence, habit formation does not
account for self-serving changes in tastes. Rationalizing choice, on the other hand,
demands a more rapid change in preferences and more flexibility in the direction
it goes.
The existence of rapid changes in preferences is supported by a sizeable body
of research showing a picture of a remarkably labile nature of preferences (see
Slovic [1991]). Illustrative of this are the following two well-known examples.
• Endowment effects: once goods are part of one’s endowment, the valuation
immediately increases sharply. This effect is present even if the subjects are
made familiar with the object on beforehand, thereby excluding learning
arguments as an explanation (Loewenstein and Adler [1995], Thaler [1980]).
• Framing: another well established phenomena is the sensitivity of choice to
the way that a choice problem is formulated. For example, the valuation of
a gamble is sensitive to whether the outcomes are framed as gains or losses
relative to the status quo (Tversky and Kahneman [1992]).
The unstable nature of preferences paves the way for manipulating preferences in
a way that is beneficial to oneself (though not necessarily consciously). Thus, not
Evidence from psychology
161
surprising on retrospective grounds, Slovic [1991] reports the existence of ”preference management”: preferences are constructed on the spot by ”...discarding
nonessential differences, adding new attributes into the problem frame in order
to bolster one alternative, or otherwise restructuring the decision problem to create dominance and thus reduce conflict and indecision.” (Slovic [1991, 500]). In
a similar vein, Bénabou and Tirole [2000] speak of ”awareness management”:
people reframe performance by remembering successes, forgetting failures, and
by trying to convince themselves that the act was not so bad. Thus, although
”... the individual updates his beliefs according to broad Bayesian principles ...
it is also widely recognized that information acquisition and belief updating are
subject to self-serving biases.” (Bénabou and Tirole [2000, 2]).
From the marketing literature it is clear that firms have acknowledged that people rationalize choice. To give two examples, according to Dibb et al. [1997, 108],
buyers seek positive information to justify their choice. They claim that ”motoring journalists often note with amusement that car shows and exhibitions are
frequented by consumers who have just recently purchased a new car”. Berkowitz
et al. [1994, 144] take the advertising campaign by Buick as an example, which
had as message ”Aren’t you really glad you bought a Buick”. Firms know that
people seek to justify their choice afterwards, and they give response to this desire.
These findings are incorporated in the model of section 7.4. For a more elaborate
exposition of evidence of preference management I refer to Bénabou and Tirole
[2000] and Rabin [1995]2 .
7.3 Evidence from psychology
The theory of cognitive dissonance roughly boils down to the idea that people
rationalize their choice. The idea is not limited to goods with social externalities.
In fact, despite its relative simplicity, there is a wide range of applications (Aronson [1988], Akerlof and Dickens [1982]). This section describes some experimental
2 Other
contributions to economics that try to explain the findings by social psychologists by formalizing the
theory of cognitive dissonance include the following. The focus of Akerlof and Dickens [1982] is on the purchase
of safety equipment in an environment of uncertainty, Akerlof [1991] considers time-inconsistency, Dickens [1986]
criminal behaviour, Rabin [1994] examines social norms, and Rabin [1995] moral behavior. James and Gutkind
[1985] apply the concept to the conditional help provided by the IMF.
Chapter 7. Optimal Subsidies with Rationalizing Agents
162
results. All of them are easily understood by assuming that people are motivated
to rationalize their choice3 .
In a very early experiment by Brehm [1956], some women were asked to rate
various household products. The experimenter selected two of the appliances that
were presented wrapped to the women, and which were rated equally by the
women. The women were then told that they could choose one of these two
products as a reward for participation (without knowing on which base these
were selected). Still wrapped, they were asked to evaluate these two products
again. A systematic feature of the second evaluation relative to the first one,
was that the chosen product increased in valuation, whereas the rejected product
apparently became less attractive to these women.
Aronson and Carlsmith [1963] designed an experiment where children were
asked to rate several toys in attractiveness and were then left alone. In one condition, the experimenter took the second rated toy with him. In a second condition,
the toy was left with the child but the child was asked not to play with it, with
the added mild threat that the experimenter would otherwise be annoyed. In the
third condition, the threat was more severe, announcing that he would be very
angry. No child played with the toy. After that, the experimenter returned and
asked for another evaluation. As it turns out, the perceived attraction increased
in conditions 1 and 2, where the child had enough external reasons not to play
with the toy. In the mild threat condition, however, the attraction was less in the
second evaluation. Apparently, the child had no really good justification for not
playing with the toy and so tried to rationalize behavior by reasoning that the
toy was not so attractive after all.
Festinger and Carlsmith [1959] report an experiment where subjects had to listen to a rather boring seminar. After the seminar, some were asked to tell the next
participants that the seminar was going to be very interesting. Some were paid $1
for this, others $20. All participants were asked to rate the seminar. Interestingly,
the subjects who were paid $1 showed a much more positive evaluation record
than the other groups. This could be expected. Some individuals were asked to
tell a lie, and nobody rejected this request. Likely, telling a lie is not congruent
with the self-image of being a decent person. This can create some dissonance.
However, the group of people which received $20 has a clear rational for lying:
3 Most
of the material in this section draws upon the expositions by Aronson [1988] and Brigham [1991].
163
Evidence from psychology
they were paid considerably for this. Hence, the only group where the lie created
dissonance was the group which only received $1 and they indeed revised their
opinion most.
If we assume that people try to rationalize behavior, this has an interesting
consequence for how people should recall things. Consider a situation where there
are good reasons pro and contra an individual’s position. From all the reasons
pro and contra, some make sense and others are less plausible. Which ones would
be best to recall for an individual? In order to rationalize the position taken, the
best thing to do is to remember the sensible reasons congruent with the position
taken, and the implausible arguments that are incongruent. In an experiment by
Jones and Kohler [1959] exactly this pattern was found.
Three more things are worthwhile to note. First, dissonance seems to bring
a real physiological arousal. An experiment by Croyle and Cooper [1983] shows
that conditions creating high-dissonance situations show more skin conduction
responses. According to them, this is a reliable indicator of physiological arousal.
This means that the phenomenon of dissonance goes beyond subjective selfreports (Aronson [1988]). Relatedly, attitude change is not a superficial tendency
of people to pretend they were changing (Baumeister [1998]).
Second, attitudes are, if properly measured, a reliable indicator of behavior
(Brigham [1991]). In a famous study by Lepper, Greene and Nisbett [1973] children were asked to draw a picture using attractive magic markers. Some children
were paid for this. In a second session, the children were left alone to play freely
with the materials. Those children who were not paid before (and had therefore
no good rational to participate other than that the magic markers were enjoyable
in themselves to play with) were more likely to spend time with the materials.
Hence, instead of measuring attitude changes, this study shows that also behavioral changes are induced. This is an important aspect from an economics point
of view, since if changed attitudes do not lead to changes in behavior they are
not very interesting from the perspective of choice theory.
Finally, the attitude changes are persistent. Between the first and second session
of the experiment by Lepper et al. [1973] was a time span of several days. In
a study by Freedman [1965] children were less likely to play with previously
forbidden toys, even after nine weeks had passed.
Chapter 7. Optimal Subsidies with Rationalizing Agents
164
7.4 A simple model
The aim of this section is to present a simple model that captures the foregoing observations. Still abstracting from preference management, subsection 7.4.1
introduces the unpleasant feeling that people experience when doing something
threatening to their self-concept into the utility function. In subsection 7.4.2 I
allow for preference management. Here, choice can be rationalized along the ways
described in the introduction and the previous section. Although this clearly is
a dynamic process, for simplicity the model is essentially static. It is my feeling
however, that the main results go through in a more general setting. Because the
ultimate aim of the chapter is to consider the effects of subsidies on efforts, this
is subsequently studied in section 7.4.3.
7.4.1 Preferences and choice
The standard utility function
Consider an individual with income I whose preferences can be described by a
quasilinear utility function. Later I explain how this specification simplifies the
analysis considerably but it is noteworthy to mention that it is not crucial to
obtain the results. Thus, let the following standard utility4 function represent her
preferences:
u(x, y) = xα + y,
0 < α < 1.
(7.1)
Here, x and y are both bundles of goods. My concern here is only the bundle
of x goods. Within the bundle of indivisible x goods a distinction can be made
between x1 (the environmental damaging good) and x2 (the green product). These
goods are in principle good substitutes but each good has its own advantages
which will have to be weighed against each other. Thus, for example, if x is
the product ’wood’, then x1 and x2 may be tropical hardwood and certificated
wood guaranteeing forest preservation, respectively. The parameter δ reflects their
relative attractiveness as perceived by the individual in the following way:
x = x1 + δx2 .
4 Standard
(7.2)
refers here to the fact that normally dissonance is not an argument in the utility function, not
to the use of a quasilinear specification.
A simple model
165
Hence,
u(x1 , x2 , y) = (x1 + δx2 )α + y.
(7.3)
For simplicity, it is assumed that the x-goods are indivisible. Note that at most
one of the x-goods will consumed. As a shortcut, where no confusion can arise I
employ the notation u1 = u(x1 , 0, y) and u2 = u(0, x2 , y). Hence, ui denotes the
utility of consuming the bundle (x, y) knowing that good xi is consumed.
The extended utility function
I now wish to incorporate the unpleasant feeling, or dissonance, into the framework. I propose the following extended utility function of an individual who
chooses product xj :
ûj = uj − dj .
(7.4)
This specification gives credit to the thought that reducing dissonance, d, increases extended utility. Hence, reducing dissonance is a motivational factor. Cognitive dissonance is measured by the relative attractiveness (in standard utility
terms) between the chosen alternative and the rejected good5 . This seems reasonable. Buying environmental damaging products creates dissonance, but less so if
their green counterparts are clearly inferior in quality. Let dj, denote dissonance
if good xj is consumed and good x−j is rejected, where j = 1, 2. Then:
dj = d(uj , u−j ),
5 Note
(7.5)
carefully the use of ’rejected’. It is possible that another good is valued positively but is out of the
budget restriction. In this sense, it is not rejected and causes no dissonance. Similarly, coerced regulations (such
as some taxes) that cannot be avoided do not cause dissonance. This is known in the psychological literature as
’forced compliance’.
Chapter 7. Optimal Subsidies with Rationalizing Agents
166
FIGURE 7.1.
with the properties:
∂d(uj , u−j )
< 0,
∂uj
∂d(uj , u−j )
> 0,
∂u−j
∂ 2 d(uj , u−j )
> 0,
∂u2j
(7.6)
∂ 2 d(uj , u−j )
> 0,
∂u2−j
and furthermore that uj = u−j implies d(uj , u−j ) = d(u−j , uj ).These conditions
are shown graphically in figure 7.1. Dissonance decreases as the chosen alternative
gets better or as the rejected alternative gets worse. Dissonance is maximal at the
point where both alternatives are equally attractive. (I therefore might as well
have labeled individuals as indifference averse.)
At this point, it is worthwhile to emphasize that the way dissonance is incorporated is not in contradiction with standard rationality assumptions. If the
standard utility function u represents the preference relation of the individual,
then, with the assumptions on (7.5), so does û.
Lemma 9 û is a monotonic transformation of u.
A simple model
167
Proof. I want to show that u1 S u2 is equivalent to û1 S û2 . Suppose first that
u1 > u2 for any given y. Since ∂d(uj , u−j )/∂uj < 0 and ∂d(uj , u−j )/∂u−j > 0
we have the following chain: d1 = d(u1 , u2 ) < d(u1 , u1 ) < d(u2 , u1 ) = d2 . Hence,
û1 = u1 − d1 > u2 − d2 = û2 . The other cases are similar.
The intuition behind this result is the following. Suppose that good x1 is preferred
to x2 . Choosing x1 causes some dissonance. But would x2 be consumed, and
consequently x1 be rejected, then dissonance would be even higher. This can also
be seen in figure 7.1. In the area left to the point u1 , good x1 is preferred to x2
(u1 > u2 ). As is clear from the figure, in this area the d2 -curve is everywhere
above the d1 -curve. This means that dissonance of rejecting x1 is higher than
dissonance of rejecting x2 . Clearly, if both standard utility of x1 is higher and
dissonance is lower, then the extended utility of consuming x1 is also higher.
This result can be used to derive the individual’s choice without specifying dj .
I assume that the individual receives a lump-sum subsidy s from the government
conditional on consumption of the green good, x2 . To simplify even more, I assume
that both bundles of goods are consumed and that all prices are unity. Finally,
v(p, I, s) defines the standard indirect utility function that gives the maximum
standard utility achievable at the prices, income and subsidies the individual
faces. It is then straightforward to show the following choice behavior (see also
figure 7.2):
Proposition 15 There exists a subsidy level s̄ ≡ k1 − k2 δ α/(1−α) such that (i)
for all s < s̄ good x1 is consumed and v = I + k1 and for all s ≥ s̄ good x2 is
α
consumed and v = I + s + δ 1−α k2 , where k1 and k2 are positive constants and (ii)
s̄δ < 0.
Proof. Let u(x, y) = xα + y and for simplicity px = py = 1. We can proceed
in two stages. First, an optimal (x, y) combination can be determined and then
the optimal (x1 , x2 ). Since it is assumed that y > 0 and marginal utility of y is
equal to 1, it must be that an additional unit of x gives a marginal utility of
less than unity. Define ϕi = min ∂u(x, y)/∂x s.th. ∂u(x, y)/∂x ≥ 1. Suppose x1 is
consumed. Then x1 is consumed up to the point where αxα−1
= ϕ1 , or:
1
x∗1 = (α/ϕ1 )1/(1−α) .
(7.7)
Consumption of good y is determined by the budget restriction:
py y ∗ = I − px x∗1 ⇔ y∗ = I − (α/ϕ1 )1/(1−α) .
(7.8)
Chapter 7. Optimal Subsidies with Rationalizing Agents
168
FIGURE 7.2.
Substitution of the demand function in the utility function gives the indirect
utility function
v(p, I, s) = v(1, I, 0) = I + k1 ,
(7.9)
where k1 ≡ (α/ϕ1 )α/(1−α) − (α/ϕ1 )1/(1−α) . (k1 > 0 for α < 1.) Next, suppose that
x2 is consumed. Then it follows that:
x∗2 = (α/ϕ2 )1/(1−α) δ α/(1−α) .
(7.10)
The budget now includes the subsidy, hence in this case py y∗ = I + s − px x∗2 .
Indirect utility is in this case given by:
v(p, I, s) = I + s + k2 δ α/(1−α) ,
(7.11)
where k2 ≡ (α/ϕ2 )α/(1−α) − (α/ϕ2 )1/(1−α) . The individual consumes good x2 if
and only if indirect utility is higher, i.e. if I + s + k2 δ α/(1−α) ≥ I + k1 and equality
at the threshold subsidy level s̄. Hence:
s̄ ≡ k1 − k2 δ α/(1−α) .
Since k2 > 0, s̄δ < 0.
(7.12)
A simple model
169
7.4.2 Rationalizing choice
The absence of any general principles or rules, either of personal or
administrative morality, which made it possible for him either to agree
or disagree with anybody according to what was wanted at the time. L.
Tolstoy, Resurrection.
Plagued by an unpleasant feeling, the individual then tries to rationalize her
choice. By focusing on certain arguments while ignoring others, she restructures
the problem as to convince herself that the choice she made was indeed the right
one. This means that if she chose good x1 she puts efforts in trying to find
arguments that bolster the choice of this good, and deprive the attractiveness of
the rejected good x2 . In case she chose good x2 , she does just the reverse6 .
In the model, the relative valuation of the two x goods is given by δ. The
psychological or physical efforts, e, are therefore reflected in a change in δ. Let
this change, ∆δ, be governed by:
∆δ = δ(γe − (1 − γ)e),
(7.13)
with γ = 1 if v1 ≤ v2 and
γ = 0 if v1 > v2 .
where δ(0) = 0 and δ 0 (·) > 0. For simplicity, it is also assumed that δ 00 (·) = 0.
The properties of γ assure that someone who chose x1 decreases δ (making x1
look relatively more attractive) and someone who chose x2 increases δ (making
x2 look relatively more attractive), so that in both cases efforts indeed rationalize
choice.
It goes without saying that, unlike the character of Tolstoy, for most people
rationalizing choice comes at a cost, c. Information has to be gathered, and rehearsed to recall later on, principles of decency have to be given up, mental efforts
have to be put into convincing oneself of one’s rightness, and so forth. These costs
put a limit to the ability to change your opinion. Hence,
c = c(e),
6 Note
(7.14)
that the timing is that first decisions are made and that they are rationalized only afterwards. This
is in accordance with the psychological literature (see for example Festinger [1957]) but is not crucial to obtain
the results in this paper.
Chapter 7. Optimal Subsidies with Rationalizing Agents
170
where c(0) = 0, and with the usual assumptions c0 (e) > 0 and c00 (e) > 0.
In sum, the individual solves the following program:
e∗ ∈ arg maxe≥0 v̂(e) − c(e),
(7.15)
where v̂ = v − d is indirect extended utility. Given the assumptions on δ(e) and
c(e) there is a unique effort level e∗ that optimizes attitudes. The next section
examines some properties of the optimal effort level in relation to the subsidy
level.
7.4.3 The effect of a subsidy
It is obvious that the optimal effort level will somehow be related to the subsidy
level. The subsidy not only influences choice behavior and standard utility by
changing the budget restriction, but ultimately also extended utility through its
effect on dissonance. One may conjecture that the efforts put in rationalizing
behavior pay off most when dissonance is most severe. In this subsection I show
this conjecture to be true.
First, consider the effect of a subsidy on the level of dissonance. A low subsidy
makes good x1 relatively attractive to purchase, and a high subsidy does the same
for good x2 . Both a low and a high subsidy in itself therefore provide a good
rationalization of choice. In effect, low and high subsidies create little dissonance.
It are the intermediate subsidy levels that create most dissonance, peaking at the
point of indifference. This leads to the following lemma:
Lemma 10
∂d(·)
∂s
> 0 for s < s̄ and
∂d(·)
∂s
< 0 for s ≥ s̄.
Proof. Since v1 is constant in s, and v2 increasing in s (see prop. 1), it follows
straightforward from the assumptions on d that d(v1 , v2 ) is increasing in s and
d(v2 , v1 ) decreasing.
This immediately leads us to the conclusion that, without preference management, up to a certain point (that is, s̄), increasing subsidies make individuals
worse off by creating dissonance. When the subsidy exceeds this level, however,
increasing subsidies makes individuals better off by enriching their consumption
level and taking away dissonance. Thus:
Proposition 16
∂v̂(·)
∂s
< 0 for s < s̄ and
∂v̂(·)
∂s
> 0 for s ≥ s̄.
A simple model
171
Proof. For s < s̄, by proposition 15 v(p, I, s) is constant in s and by lemma 10
dissonance is increasing in s. Hence, v̂(p, I, s) is decreasing in s. For s ≥ s̄, by
proposition 15 v(p, I, s) is increasing in s (see prop. 15) and (by lemma 10) d is
decreasing in s. Hence, v̂(p, I, s) is increasing in s.
The existence of dissonance depresses extended utility. The threshold level s̄ is
the point where dissonance is most severe. It is also the point where the individual
gains most from rationalizing her choice. Dissonance decreases at subsidy levels
that are further away from this threshold, and so do the gains from rationalization.
In terms of optimal efforts, this means that efforts are increasing in the subsidy
level up to s̄, whereas for any higher subsidy level efforts are decreasing:
Proposition 17
∂e∗
∂s
> 0 if s < s̄ and
∂e∗
∂s
< 0 if s ≥ s̄.
Proof. Let v̂ 0 (e(s), s) ≡ v̂(e(s), s) − c(e(s)). By definition:
¯
∂v̂ 0 (·) ¯¯
≡ 0.
∂e ¯e=e∗
Therefore at e∗ :
(7.16)
∂ 2 v̂ 0 (·) ∂e ∂ 2 v̂ 0 (·)
+
≡ 0,
∂e2 ∂s
∂e∂s
(7.17)
∂e
∂ 2 v̂0 (·)/∂e∂s
=− 2 0
.
∂s
∂ v̂ (·)/∂e2
(7.18)
which is identical to:
Since v̂ 0 (·) is maximal at e∗ , the second-order condition already requires that
∂ 2 v̂0 (·)/∂e2 < 0. Consider first the case where s < s̄. Define δ̃ ≡ δ α/(1−α) . We have
that:
∂e
∂ 2 v̂ 0 (·)
∂ 2 d(uj , u−j ) 0
k δ̃ (e).
(7.19)
= sign
= sign −
sign
∂s
∂e∂s
∂u2−j
2 0
v (·)
The last equality follows from the fact that ∂∂e∂s
= 0. Since γ = 0 we have that
2
0
∗
∂ d(uj ,u−j )
δ̃ (e) < 0 and because
> 0 we have that ∂e
> 0 as stated. The case
∂s
∂u2
−j
0
where s ≥ s̄ is similar but now γ = 1 so that δ̃ (e) > 0, and now
a sufficient condition for
∂e∗
∂s
∂ 2 d(uj ,u−j )
∂u2j
> 0 is
< 0.
Proposition 17 has the following consequences for the attitudes held. For low
subsidy levels (s < s̄) dissonance can be reduced by decreasing δ. The higher
the subsidy, the more efforts are made and, consequently, the larger the attitude
change. For high subsidy levels (s ≥ s̄) it is optimal to increase δ as compared
Chapter 7. Optimal Subsidies with Rationalizing Agents
172
FIGURE 7.3.
to the initial beliefs. The higher the subsidy, the smaller the increase however. In
this range there is an inverse relationship between the subsidy and the attitude
change δ. This result is in accordance with much of the psychological literature
(see for example Aronson [1988] and section 7.3).
Figure 7.3 shows the implications of proposition 17 for the standard utility
function. The solid line v2 represents standard utility at any given subsidy level.
The dotted lines show the magnitude of the shift of the v2 -curve. For example,
at point s̃, the standard utility function is shifted by the distance f so that the
new standard utility function of consuming x2 is given by v20 .
Although strictly speaking the model is a static one, proposition 17 still gives
some hints to interesting dynamic implications. Suppose that the objective of
the government is to stimulate the consumption of green products. It may in
principle do so by providing any subsidy greater or equal than s̄. But proposition
17 suggests that the government does best by giving a subsidy exactly equal to
s̄. In this way, the individual is stimulated most to change his attitude towards
the green product. The dynamic consequence is that this subsidy has the highest
chance of inducing persistent consumption of green products after removal of the
subsidy. A subsidy level of s̃, for instance, induces a shift of the v2 -curve by the
distance f, enough to make the individual prefer the green product even at a
Discussion
173
subsidy level of zero (see Figure 7.3)7 . On the other hand, a higher subsidy level
like ŝ does nothing to the attitude of individuals and removing the subsidy means
that the individual switches back to consumption of the environmental damaging
good. In sum, a high subsidy is ’too much of a good thing’: it affects current choice,
but it fails to affect attitudes and therefore choice at any future time period. Note
also that according to the same logic, subsidies that are below the threshold level
may have undesired consequences in that people try to rationalize their choice
by degrading the green good. This is especially the case for subsidies near the
threshold level. Seen from a dynamic perspective, this means that it is better not
to subsidize at all than to subsidize a bit. The point is that the stimulation of
green goods can become unnecessarily very costly if the government fails to take
into account changes in attitudes. Subsidies that are not very well targeted are
costly: in a static as well as in a dynamical sense.
7.5 Discussion
Robustness
Proposition 17 is the main result of the chapter and it is therefore interesting to
see whether or not it can be generalized to other than quasilinear specifications
of the utility function. The interesting part of the proposition is that for subsidy
levels exceeding s̄, where there is a negative relationship between subsidies and
efforts. The focus is therefore on this range of subsidies.
As usual, the sign of the change in effort as a response to a change in the
subsidy is determined by the sign of the cross partial derivative of the objective
function; ∂ 2 (v̂(e, s) − c(e))/∂e∂s. In general, with v̂ = v − d this is given by:
· 2
¸
∂ 2 (v̂(e, s) − c(e))
∂ v
∂ 2d
sign
= sign
−
.
(7.20)
∂e∂s
∂e∂s ∂e∂s
Note that the first term measures the effect of a subsidy on the marginal benefits
of efforts. In general, this effect is positive: at higher subsidy levels, the individual
7 It
is fairly easy to show that attitudes can persist after withdrawal of the subsidy. Consider an individual
that faces subsidy level s and suppose that she makes efforts to increase δ to δ0 . By proposition 15, the new
threshold subsidy level is s0 < s. Hence, there exists a range of subsidies smaller than s for which the individual
continues to consume x2 .
Chapter 7. Optimal Subsidies with Rationalizing Agents
174
spends more on good x2 and this increases the marginal benefits of efforts. Under
a quasilinear specification, however, all additional income is spent on the y-good
and the effect drops out, strengthening proposition 17. Note furthermore that if
an inverse relationship between efforts and subsidies is indeed observed, then it
can only attributed to the dissonance term. Without dissonance, subsidies always
work as a reinforcer of making efforts through the term ∂ 2 v/∂e∂s.
Crowding-out
Crowding-out is an interesting phenomenon that is supported by a substantial
amount of empirical and experimental evidence (see e.g. Deci and Ryan [1985],
Frey [1997], Kohn [1993], and Chapters 1, 3, and 6). Crowding-out is said to occur
at instances where money has a perverse effect on motivation. That is, rather than
being encouraging, higher rewards reduce the motivation to undertake a activity.
The common interpretation of this effect is that, although higher rewards create
more external motivations to undertake an activity, at the same time it tends to
destroy the intrinsic motivation people have.8
The reduction in intrinsic motivation has sometimes been given a dissonance
reduction interpretation.9 Seen from that perspective, sufficiently high rewards
provide enough external justification to perform a task, whereas relatively low
rewards can only justify efforts if people can convince themselves that the task
is intrinsically motivating. Intrinsic motivation is in the latter case likely to be
built up, whereas in the former case people feel no need to do so. Higher rewards
thus result in lower intrinsic motivation than lower rewards, mimicking the result
of the previous section (proposition 17). This result notwithstanding, cognitive
dissonance theory cannot explain true crowding-out. Higher rewards may give
less incentives to build intrinsic motivation, it does not give incentives to destroy
it. If anything, one would expect that higher rewards give more justification to
undertake the rewarded activity, even if insignificantly so (see however the next
section for an explanation why higher rewards may undermine motivation after
all).
8 In fact, crowding-out is not confined to situations where rewards are provided. Other external motivations,
such as punishments, can produce the same result.
9 See for example Dickens [1986] who presents a formal model where an increase in punishments can lead to
more criminal activity. His explanation is based on dissonance reduction.
Discussion
175
Alternative explanations
It should at this point be noted that many of the experimental results put forward by psychologists in fact do have an appealing more traditional economic
interpretation, namely, that the rewards are signals. To continue with the leading
example of this chapter, suppose that the quality of the green good is known to
be relatively low. This means that a relatively high subsidy is needed to make the
green good equally attractive as its environmental damaging counterpart. If, on
the other hand, the consumers cannot directly perceive the quality of the green
good, they may infer from a high subsidy that the quality must be low. In a
related paper, Bénabou and Tirole [2002] present a formal model where a higher
bonus signals a more difficult task or lower ability to the agent who has to perform the task (see also Chapter 6). Under some conditions, this can lead to the
same predictions as proposition 17. High subsidies signal a low quality, and once
removed, the consumer is no longer willing to pay more for the green good. Low
subsidies may signal a high quality and result in permanent consumption of the
green good, even after withdrawal of the subsidy.
The reward as a signal is a reasonable alternative view, and it plays without
doubt a role in many situations. There are, however, experiments where a clear
signal is lacking and yet attitude changes still occur. For instance, in one study
children were offered an opportunity to cheat but no payments were made whatsoever. They became more lenient towards cheating when they did not resist the
temptation (see Aronson [1988]). In the experiment from section 7.3 where women
were asked to rate several appliances, no new information was revealed during
the experiment because the rewarded appliance was given wrapped to them until
after their second rating. In the experiment by Lepper et al. [1973] where children
were given the task to draw a picture, it was told to the children that a reward
would follow just for drawing a picture, independent of the endresult. These rewards therefore carry no clear-cut information about their individual ability or
task difficulty. The result of decreased interest was replicated in a follow-up study
by Greene and Lepper [1974] where it was told to the children that all of them
would receive a reward for doing the task, again revealing no information. Finally, the experiment by Jones and Kohler [1959] where people were asked which
arguments they recalled, it is not obvious what kind of signal could have been
given.
Chapter 7. Optimal Subsidies with Rationalizing Agents
176
A competing alternative interpretation from psychology to cognitive dissonance theory is self-perception theory, developed by Bem [1972]. This theory,
in essence, reverses the causality between attitudes and behavior. Self-perception
theory states that people infer their attitude from their behavior (see also chapter 1 and section 2.3.5 of chapter 2 on self-signalling). Thus, the people in the
experiment of Festinger and Carlsmith [1959] who received $1 for telling the next
participants that the seminar is going to be a lot of fun, must have reasoned that,
because they did this for just one dollar, they really must have liked the seminar.
Even though this theory has many merits in many situations, it gives in many
instances ultimately the same behavioral predictions, and in some cases it is
somewhat less convincing as an explanation. First, the theory assumes imperfect
information about one’s self, which is more reasonable for people who try to
quit smoking for the first time than for people who have already tried many
times before. It therefore depends on whether one believes that the people were
capable of valuing the seminar directly or had to infer this from their behavior.
Since the participants in this experiment were college men, one can safely assume
that they were capable of directly assessing the seminar10 . Furthermore, selfperception theory, as well as the signalling approach discussed above, assumes
that no physiological arousal takes place in any condition. However, as pointed
out earlier, such a physiological arousal has been measured to be present.
7.6 Conclusions
While psychologists have put great efforts in understanding the formation of
preferences in order to explain changes in behavior, the economic approach insists
on explaining any changes in behavior by changes in income, relative prices, and
information. The purpose of this chapter is to show some of the consequences it
can have when attitude changes have been taken care of as well. Under, in my
view, fairly intuitive assumptions, I have been able to replicate the experimental
finding that low rewards induce more attitude change than high rewards.
10 It
is not unreasonable that self-perception theory becomes more relevant if a significant time period has
elapsed between the action and evaluation. Since retrospective evaluations of past utilities are known not always
to be reliable (Kahneman [1994]), inferring preferences from past behaviour may be more accurate than from
recalled utilities. I do not know of any study which has worked this out.
177
Conclusions
The focus of this chapter has been on a feeling of dissonance to explain the
data. Others have stressed the informational aspects. Likely, both elements are
present: informational signals and an unpleasant physiological arousal. Unfortunately, in many cases the literature seems insufficiently conclusive in establishing
the relative weight of informational aspects. In my view, the results of this chapter should be seen as complementary to a signalling interpretation. The latter
approach is taken up in Chapter 6.
8
Summary in Dutch
Beknopte inleiding in de onderwerpen
Vaak wordt er in economische modellen aangenomen dat mensen egoïstisch zijn en
alleen om geld en goederen geven. In veel gevallen kunnen op basis van deze veronderstellingen goede verklaringen geboden worden voor hoe mensen zich gedragen.
Maar er blijven ook veel onopgeloste vraagstukken over: waarom laten mensen
fooien achter? Waarom zijn ze soms minder gemotiveerd om iets te doen wanneer
er een beloning tegenover staat? Waarom verzamelen mensen zoveel informatie
nadat ze een aankoop hebben gedaan in plaats van dat vooraf te doen? En waarom
kopen mensen Max Havelaar koffie? Deze en andere vragen stel ik centraal in dit
proefschrift.
Om een antwoord te geven op bovenstaande vragen verwerk ik onderzoek uit
de psychologie in economische modellen. Zo stel ik bijvoorbeeld in navolging van
de psychologische literatuur dat mensen sociale waardering willen krijgen voor
hun gedrag. Andere concepten die ik overneem zijn bijvoorbeeld dat mensen hun
gedrag graag rationaliseren en dat ze onvolledige informatie over hun eigen persoonlijkheid hebben. Door economische modellen uit te breiden met dit soort
ideeën kan een verklaring gegeven worden voor veel gedrag dat afwijkt van standaard economische modellen.
Het proefschrift is ingedeeld in twee centrale thema’s. Het eerste thema gaat
in op de vraag waarom mensen aan elkaar of aan een liefdadigheidsinstelling
Chapter 8. Summary in Dutch
180
geven. Het tweede thema betreft hoe mensen reageren op beloningen. Eerst ga ik
dieper in op deze twee thema’s. Vervolgens geef ik een inhoudelijk overzicht van
de individuele hoofdstukken.
Geven
De meeste mensen geven een behoorlijk bedrag weg aan familie, vrienden,
of liefdadigheidsinstellingen. Het gaat hier om uiteenlopende dingen zoals verjaardagscadeaus en bloed, maar ook vrijwilligerswerk kan opgevat worden als een
gift. Men is geneigd om te denken dat het hier om vrijwillige en vrijblijvende
schenkingen gaat, maar niets is minder waar. In de werkelijkheid blijkt er een
sterke sociale druk te bestaan om iets terug te geven nadat men iets ontvangen
heeft. De ontvanger heeft als het ware een schuld uit staan die terug betaald moet
worden. Het is dan ook niet helemaal verwonderlijk dat vergif de tweede betekenis
van gift is.
Vanuit een economisch perspectief is het in eerste instantie verwonderlijk dat er
van alles weggegeven wordt, en slechts een klein deel daarvan in de vorm van geld.
Volgens de micro-economie kan geld nooit slechter zijn dan een cadeau. Immers,
de ontvanger kan met dat geld hetzelfde cadeau kopen, of iets wat hij nog liever
heeft. Maar als de ontvanger beter af is met geld, waarom geven we dan zo vaak
cadeaus?
Beloningen
Het tweede thema betreft hoe mensen reageren op beloningen. Doorgaans wordt
in de economie verondersteld dat een beloning in het vooruitzicht mensen motiveert. De reden hiervan is dat de nadruk meestal ligt op het directe effect van
beloningen. Mensen geven om geld, dus het ligt voor de hand dat een beloning
hen stimuleert om harder te werken. Er zijn echter ook indirecte effecten, soms
met een tegengestelde werking. Dit blijkt uit experimenten die door psychologen
zijn gedaan.
Een deel van die experimenten meet de zogeheten intrinsieke motivatie van
mensen. Met intrinsieke motivatie wordt bedoeld dat mensen gemotiveerd zijn
om bepaalde dingen te doen zelfs als er geen beloning tegenover staat of als ze
niet onder controle staan. Een goed voorbeeld zijn kinderen die ijverig werken aan
een puzzel. Het blijkt echter dat deze intrinsieke motivatie soms afneemt zodra een
181
beloning wordt verstrekt. De kinderen gaan in eerste instantie nog harder aan de
slag, maar zodra de beloning weggenomen wordt zijn ze minder gemotiveerd dan
voorheen. Een indirect effect van beloningen is dus het verdringen van intrinsieke
motivatie. Andere experimenten laten zelfs zien dat de motivatie direct afneemt
na het invoeren van een beloning. Zo nam het aanbod van bloeddonoren af nadat
een compensatie werd verstrekt, en kinderen wisten minder geld op te halen bij
een collecte nadat ze een kleine vergoeding kregen. Er zijn dus ’verborgen kosten
van beloningen’.
Er is nog een reeks experimenten die een onverwacht effect van beloningen laat
zien. Het gaat hierbij om de invloed van beloningen op voorkeuren van mensen. In
de economie wordt verondersteld dat voorkeuren vastliggen. Psychologen hebben
echter gevonden dat mensen iets meer gaan waarderen wanneer ze er een kleine
beloning voor krijgen, terwijl dit effect niet gevonden wordt bij een hoge beloning.
Zo woonden sommige mensen een uitgesproken saai seminar bij. Er werd hen
gevraagd de volgende groep luisteraars te zeggen dat het allemaal zeer interessant
zou worden. Een deel van hen kreeg daarvoor een kleine beloning, een ander deel
een flinke beloning. Daarna werden ze gevraagd te zeggen wat ze er zelf van
vonden. Diegenen met een hoge beloning waardeerden het seminar laag, terwijl
diegenen met een lage beloning het hoog waardeerden.
Overzicht van de hoofdstukken
Het voorgaande is een beknopt overzicht van de verschillende thema’s uit dit
proefschrift. De thema’s worden dieper uitgewerkt in de hoofdstukken zelf. Het
proefschrift begint met een algemene inleiding in het onderwerp psychologie en
economie. De overige hoofdstukken zijn ruwweg ingedeeld naar de twee thema’s.
Hoofdstukken 2 tot en met 5 gaan in op de vraag waarom mensen geven. De
nadruk in hoofdstukken 3, 6 en 7 ligt op de effecten van beloningen. Dit onderscheid is uiteraard enigszins kunstmatig, want in brede zin zijn beloningen soms
ook giften, en andersom. Hieronder volgt een overzicht van de inhoud van de individuele hoofdstukken, zonder diep in te gaan op de modellen en verklaringen.
Het is vooral bedoeld als leidraad.
Hoofdstuk 2. In dit hoofdstuk worden redenen uiteengezet waarom mensen
geven. Er zijn veel redenen denkbaar: altruïsme, ruil, rechtvaardigheid, signalen
geven en sociale waardering komen allen aan bod. Echter, niet allen zijn even
Chapter 8. Summary in Dutch
182
geloofwaardig als verklaringen. Zoals gezegd zijn er twee bijzonderheden aan
geefgedrag: bijna altijd wordt iets teruggegeven (reciprociteit) en bijna nooit in
geld (inadequaatheid). Met deze twee kenmerken in gedachten wordt de verklaringskracht van iedere theorie beoordeeld.
Een eerste voor de hand liggende verklaring is dat mensen een ruil beogen met
het geven van cadeaus aan elkaar. Dit kan inderdaad een goede verklaring zijn mits
mensen geduldig genoeg zijn, want ze krijgen bijna nooit meteen een cadeau terug.
Altruïsme lijkt aannemelijk, maar kan niet verklaren waarom mensen zo zelden
geld geven in plaats van cadeaus. Het verklaart dus maar een klein deel van alle
cadeau’s. Een andere verklaring is rechtvaardigheid. Maar er wordt aannemelijk
gemaakt dat dit niet alles kan verklaren, want in sommige experimenten handelen mensen in strijd met de veronderstelde rechtvaardigheidstheorieën. De vraag
naar sociale waardering kan deze experimenten wel verklaren, en ook waarom
giften inadequaat zijn en bovendien waarom er reciprociteit bestaat (zie vooral
hoofdstuk 3). Een andere, wellicht minder voor de hand liggende verklaring, is
dat mensen iets willen signaleren met hun geefgedrag, bijvoorbeeld hoe rijk ze
zijn of hoe goed ze de ontvanger kennen. Het kan zelfs zo zijn dat iemand iets aan
zichzelf laat zien, bijvoorbeeld dat hij rechtvaardig is omdat hij fooien achterlaat
terwijl hij net zo goed het restaurant had kunnen verlaten zonder fooi.
Het hoofdstuk eindigt met een discussie waarin ik pleit voor een hybride verklaring. Bijvoorbeeld: mensen geven niet omdat ze eerlijk zijn maar omdat ze
eerlijk willen lijken en daar sociale waardering voor krijgen. Verder beargumenteer
ik dat het belangrijk is om te weten met welk doel mensen geven voordat een
institutie ontworpen wordt. Zo kan het bijvoorbeeld averechts werken om mensen
te belonen voor hun gift als ze dit doen om sociale waardering te krijgen.
Hoofdstuk 3 borduurt voor op de gedachte dat mensen geven teneinde sociale waardering te krijgen. Allereerst probeer ik aan te tonen dat mensen om
sociale waardering geven en dat hier een element van status in zit: mensen willen
vooral meer gewaardeerd worden dan hun buren. Op basis hiervan verklaar ik
reciprociteit en adequaatheid. Reciprociteit volgt uit het status effect: Wanneer
iemand geeft krijgt hij daarvoor waardering en dit motiveert de ontvanger om ook
te geven teneinde ook die waardering te krijgen. De eerste gever krijgt het liefst
meer waardering dan de ontvanger. Door geen geld te geven maar cadeaus, maakt
hij het duurder voor de ontvanger om iets terug te geven en blijft hij voorop
lopen in de race om status. Bovendien verklaart deze theorie waarom mensen
183
meer geven als ze hiervoor publiekelijk bedankt worden, zoals vaak het geval is
bij liefdadigheid. Door de publiekelijke bekendheid die eraan gegeven wordt is het
mogelijk om sociale waardering te krijgen, zelfs als de directe ontvanger anoniem
blijft.
Er valt een interessante relatie te leggen tussen de theorie uit dit hoofdstuk en
die van averechtse effecten van beloningen. Er wordt sociale waardering gegeven
voor een gift omdat een gift een opoffering is voor de gever. Het belonen van
giften maakt de opoffering kleiner, en bijgevolg ook de sociale waardering. Een
compensatie voor geven kan op deze manier averechts werken.
Vervolgens wordt in hoofdstuk 4 geefgedrag bekeken vanuit een meer macroeconomisch perspectief. Wanneer men veronderstelt dat het geven en wedergeven
een ruil ten doel heeft, dan lijkt het aannemelijk dat op termijn het marktmechanisme al het geefgedrag zal verdringen. Naarmate de markt groeit in omvang
wordt deze efficiënter en wordt geven een slechter alternatief.
Echter, in dit hoofdstuk wordt beargumenteerd dat geefgedrag niet alleen als
ruilmechanisme dient, maar ook symbolische waarde heeft. Hoofdstuk 3 ging
eerder al in op de gedachte dat een gift sociale waardering teweegbrengt. Deze
symbolische waarde komt niet tot stand via het marktmechanisme omdat deze
een betrekkelijk anoniem karakter heeft. Hierdoor wordt geefgedrag niet in zijn
geheel verdrongen. Het model laat tevens zien dat geefgedrag kan blijven bestaan
ondanks dat de markt efficiënter zou zijn geweest. Bovendien kan het voorkomen
dat de markt al het geven verdringt terwijl geven meer efficiënt is.
In de voorgaande hoofdstukken is er beargumenteerd dat er om verschillende redenen nut ontleend wordt aan geefgedrag. In hoofdstuk 5 worden de consequenties
hiervan nader beschouwd. Er wordt verondersteld dat mensen graag een meerprijs
willen betalen voor goederen die geproduceerd zijn met behulp van technieken die
sociale externaliteiten verminderen. Met sociale externaliteiten worden bijvoorbeeld productiemethoden bedoeld die schadelijke effecten op het milieu hebben,
maar ook die gebruik maken van kinderarbeid of een onrechtvaardig laag loon
betalen. De meerprijs voor producten met minder sociale externaliteiten kan als
een gift aan de werknemers of aan de maatschappij opgevat worden.
Problematisch is dat consumenten niet kunnen beoordelen welke techniek de
producent gebruikt heeft. Ze kunnen bijvoorbeeld niet zien of er gebruik is gemaakt
van kinderarbeid. Daarom zullen ze niet bereid zijn om een meerprijs te betalen.
Chapter 8. Summary in Dutch
184
Op hun beurt zijn producenten niet bereid te investeren in productiemethoden
met minder sociale externaliteiten.
Er worden twee manieren bekeken om dit probleem op te lossen: een standaard
invoeren, of labels invoeren. Wanneer de overheid een standaard op de productiemethode invoert, dan betekent dat dat alle consumenten gedwongen zijn om
het product tegen die standaard te kopen, of om niets te kopen. Wanneer labels of
certificaten worden ingevoerd dan krijgt de producent (en daarmee de consument)
de vrijheid om te produceren volgens hun eigen gekozen productietechniek of volgens de techniek waarvoor ze een certificaat krijgen. De consumenten met een hoge
betalingsbereidheid zullen het gecertificeerde goed kopen, die met een lagere betalingsbereidheid het ongecertificeerde goed. Vervolgens wordt bekeken welke van
de twee mogelijkheden sociaal optimaal is. Er wordt gevonden dat een certificaat
beter is voor een interval van consumentenheterogeniteit. Wanneer consumenten
of relatief homogeen zijn of relatief heterogeen, dan is een standaard welvaartsmaximaliserend.
Hoofdstuk 6 bestudeert de effecten van beloningen op zelfvertrouwen in een
principaal-agent model. Het model is een uitbreiding op de literatuur die laat
zien dat een bonus een signaal kan zijn dat het een moeilijke taak betreft of
dat de principaal de talenten van de agent laag inschat. Dit verklaart waarom
beloningen negatieve consequenties kunnen hebben. In die literatuur is de focus
gericht op beloningen zoals vastgelegd in een contract. In dit hoofdstuk wordt
de aanname gemaakt dat de uitkomst alleen geobserveerd kan worden door de
principaal. Dit maakt een contract onmogelijk. Er wordt dus gekeken naar het
effect van onverwachte beloningen.
Dit hoofdstuk biedt een mogelijke verklaring waarom onverwachte beloningen
in eerste instantie gegeven worden. Een cruciale veronderstelling hierbij is dat de
principaal over meer informatie beschikt dan de agent wat betreft succesvolheid.
De principaal weet of de opdracht een succes is of niet, terwijl de agent slecht een
idee hierover kan vormen. De theorie heeft dus betrekking op situaties waar de
agent nog in een leerfase zit: een kind die piano leert spelen of een werknemer die
pas begonnen is aan zijn nieuwe baan. Een andere mogelijkheid is dat de agent
geen overzicht op het geheel van activiteiten omdat hij zich specialiseert op een
klein onderdeel ervan, terwijl de principaal het totale proces kan overzien.
In deze opzet kan een onverwachte beloning een signaal zijn dat de agent het
tot een succes heeft weten brengen. Een beloning is dus goed nieuws. Dit schroeft
185
het zelfvertrouwen van de agent omhoog, die op zijn beurt de volgende periode
meer gemotiveerd is. We contrasteren dit met de literatuur die laat zien hoe een
bonus slecht nieuws kan geven. We geven een conditie aan wanneer een beloning
goed nieuws is. Deze conditie sluit goed aan bij de vele experimentele resultaten.
In het laatste hoofdstuk, hoofdstuk 7, wordt er gekeken naar veranderingen
in preferenties. Terwijl de prikkels om gedragsveranderingen teweeg te brengen
uitgebreid bestudeerd zijn, is er weinig aandacht geweest binnen economie naar
veranderingen in voorkeuren. Sociaal psychologen hebben daarentegen veel aandacht besteed aan de vorming van preferenties, waarbij ze hebben proberen aan
te tonen dat voorkeuren niet stabiel zijn. Voorkeuren lijken inderdaad niet stabiel
te zijn, maar in veel gevallen wel voorspelbaar. In dit hoofdstuk worden prikkels
en veranderingen in voorkeuren onderzocht. Veel van de veranderingen kunnen
verklaard worden door aan te nemen dat mensen hun keuze achteraf rationaliseren.
Het basisidee is dat mensen er een onaangenaam gevoel bij krijgen (cognitieve
dissonantie) wanneer hun geloof en gedrag niet consistent zijn met elkaar (bijvoorbeeld, je rookt terwijl dat je gelooft dat het slecht is). Om van dit onaangename
gevoel af te komen kun je je gedrag proberen te rationaliseren, dus jezelf ervan te
overtuigen dat je de goede keuze hebt gemaakt. Dit kan bijvoorbeeld door informatie te zoeken die overeenkomt met je gedrag, en informatie te vermijden die er
niet mee in overeenstemming is.
Als toepassing wordt in dit hoofdstuk speciale aandacht besteed aan consumptiegoederen met minder sociale externaliteiten, zoals milieuvriendelijke producten
(zie ook hoofdstuk 5). Het belangrijkste resultaat is dat een lage subsidie op dit
soort goederen een positieve verandering in de voorkeur ervoor teweegbrengt,
maar een hoge subsidie niet. Dit komt overeen met resultaten uit experimenten.
De conclusie is dan ook dat hoge subsidies teveel van het goede zijn: ze beïnvloeden het gedrag op dat moment maar niet dat van in de toekomst.
References
[1]
Ahlbrecht, M., and M. Weber (1997), Preference for gradual resolution of
uncertainty, Theory and Decision 43, 167-185.
Ainslie, G. (1991), Deviation of “Rational” Economic Behavior from Hyperbolic Discount Curves, American Economic Review 81 (2), 334-340.
Akerlof, G. (1970), The Market for "Lemons": Quality Uncertainty and the
Market Mechanism, Quarterly Journal of Economics 84 (3), 488-500.
Akerlof, G. (1982), Labor Contracts as Partial Gift Exchange, Quarterly Journal of Economics 97, 543—569.
Akerlof, G. (1991), Procrastination and Obedience, American Economic Review 81 (2), 1-19.
Akerlof, G. (1997), Social Distance and Social Decisions, Econometrica 65,
1005—1027.
Akerlof, G. and W. Dickens (1982), The economic consequences of cognitive
dissonance, American Economic Review 72 (3), 307-319.
Andreoni, J. (1989), Giving with Impure Altruism: Applications to charity and
Ricardian Equivalence, Journal of Political Economy 97 (6), 1447-1458.
References
188
Andreoni, J. (1990), Impure Altruism and Donations to Public Goods: A Theory of Warm-Glow Giving, The Economic Journal 100, 464—477.
Andreoni, J. (1993), An Experimental Test of the Public-Goods Crowding-out
Hypothesis, American Economic Review 83 (5), 1317-1327.
Andreoni, J. (2001), The Economics of Philantropy, in N. Smeltser and P.
Baltes (eds.), International Encyclopedia of Social and Behavioural Sciences, Elsevier Pergamon, Oxford.
Andreoni, J., and R. Petrie (2000), Social Motives to Giving: Can these Explain Fund Raising Institutions?, Mimeo, University of Wisconsin Madison.
Aristotle (1994), The Nicomachean Ethics, translated by H. Rackham, Harvard
University Press: Harvard.
Arnott, R., and J. Stiglitz (1991), “Moral Hazard and Nonmarket Institutions: Dysfunctional Crowding Out or Peer Monitoring?” American Economic Review 81, 179—190.
Arnsperger, C. (2001), Gift-giving Practices and Non-contextual Habitus, in
Gifts and Interests, A. Vandevelde (ed.), Peeters.
Aronson, E. (1988), The Social Animal, W.H. Freeman and Company, New
York, 5th Edition.
Aronson, E., Turner, J. and M. Carlsmith (1963), Communicator credibility and communicator discrepancy as determinants of opinion change, Journal of Abnormal and Social Psychology 67, 31-36.
Arrow, K. (1972), Gifts and Exchanges, Philosophy and Public Affairs 1 (4),
343-362.
Arrow, K. (1997), Invaluable Goods, Journal of Economic Literature XXXV,
757-765.
Axelrod, R. (1984), The Evolution of Cooperation, Basic Books.
Baker, G., R. Gibbons, and K. Murphy (1994), Subjective performance measures in optimal incentive contracts, Quarterly Journal of Economics CIX
(4), 1125-1156.
189
References
Bansal, S. (2002), Environmental Regulation in the Presence of Environmentally Conscious Consumers, Ph.D. thesis Indian Statistical Institute, Delhi.
Barkow, J. (1989), Darwin, Sex, and Status, University of Toronto Pess, Toronto.
Baumeister, R. (1998), The Self, in: D. Gilbert, S. Fiske, and G. Lindzey (eds.),
The Handbook of Social Psychology, McGraw-Hill, New York.
Baumol, W. and W. Oates (1988), The Theory of Environmental Policy, Cambridge University Press, Cambridge, 2nd ed.
BBC News (2002), Diamonds: A rebel’s best friend, http://news.bbc.co.uk/1/hi
/world/africa/745194.stm
Becker, G. (1974), A Theory of Social Interactions, Journal of Political Economy, 82, 1063—1093.
Becker, G. (1996), Accounting for Tastes, Harvard University Press.
Bem, D. (1972), Self-Perception Theory, in L. Berkowitz (ed.), Advances in experimental social psychology vol. 6, academic Press, New York.
Bénabou, R., and J. Tirole (2001), Willpower and Personal Rules, Mimeo
November.
Bénabou, R., and J. Tirole (2002a), Intrinsic and Extrinsic Motivation, forthcoming in Review of Economic Studies.
Bénabou, R., and J. Tirole (2002b), Self-Confidence and Personal Motivation, Quarterly Journal of Economics 117(3), 871-915.
Berkowitz, E., R. Kerim, S. Hartley and W. Rudelius (1994), Marketing,
Irwin (4th ed.).
Bernheim, B., and O. Stark (1988), Altruism within the family reconsidered:
do nice guys finish last?, American Economic Review 78 (5), 1034-1045.
Binmore, K. (1998), Game Theory and the Social Contract, vol. I, MIT Press.
Bjørner, T., L. Hansen, C. Russell, and T. Olsen (2002), The Effect of the
Nordic Swan Label on Consumers’ Choice, AKF Forlaget.
References
190
Bolle, F. (2000), Why to buy your darling flowers: on cooperation and exploitation, Theory and Decision, forthcoming.
Bolton, G. and E. Katok (1998), An experimental test of the crowding out
hypothesis: The nature of beneficent behavior, Journal of Economic Behavior & Organization 37, 315-221.
Bolton, G. and A. Ockenfels (1998), Strategy and Equity: An ERC-analysis
of the Güth-van Damme game, Journal of Mathematical Psychology 42,
215-226.
Bolton, G., and A. Ockenfels (2000), ERC: A Theory of Equity, Reciprocity,
and Competition, American Economic Review, 90, 166—193.
Bowles, S. (1998), Endogenous Preferences: The Cultural Consequences of Markets and Other Economic Institutions, Journal of Economic Literature,
XXXVI, 75—111.
Brehm, J. (1956), Postdecision changes in the desirability of alternatives, Journal of Abnormal and social Psychology 52, 384-389.
Brigham, J. (1991), Social Psychology, Harper Collins Publishers, 2nd ed.
Camerer, C. (1988), Gifts as Economic Signals and Social Symbols, American
Journal of Sociology, 94, S180—S214.
Camerer, C., L. Babcock, G. Loewenstein, R. Thaler (1997), Labor Supply of New York City Cab Drivers: One Day at a Time, Quarterly Journal
of Economics 112 (2), 407-441.
Caplow, T. (1982), Christmas Gifts and Kin Networks, American Sociological
Review, 47, 383—392.
Carmichael, H. and W. MacLeod (1997), Gift giving and the evolution of
cooperation, International Economic Review 38 (3), 485-509.
Carrillo, J., and T. Mariotti (2000), Strategic ignorance as a self-disciplining
device, Review of Economic Studies 67, 529-544.
Charness, G. and M. Rabin (2002), Understanding Social Preferences with
Simple Tests, Quarterly Journal of Economics, forthcoming.
191
References
Charness, G., G. Frechette, and J. Kagel (2001), How Robust is Laboratory Gift Exchange, Mimeo.
Cho, I., and D. Kreps (1987), Signalling games and stable equilibria, Quarterly Journal of Economics 102, 179-221.
Codere, H. (1950), Fighting with property, Unversity of Washington Press.
Coleman, J. (1990), Foundations of Social Theory, Belknap Press: Cambridge,
MA.
Cooper, R., D. de Jong, R. Forsythe, and T. Ross (1996), Cooperation Without Reputation, Experimental Evidence from Prisoner’s Dilemma Games,
Games and Economic Behavior, 12, 187—218.
Cooper, B., C. Garia-Penalosa, and P. Funk (2001), Status Effects and Negative Utility Growth, The Economic Journal 111 (473),.642-665.
Corneo, G., and O. Jeanne (1997) Conspicuous consumption, snobbism and
conformism, Journal of Public Economics 66 (1), 55-71.
Croyle, R. and J. Cooper (1983), Dissonance Arousal: Physiological Evidence,
Journal of Personality and Social Psychology 45, 782-791.
Dasgupta, P. (2000), Economic Progress and the Idea of Social Capital,325—
424 in: P. Dasgupta and I. Serageldin (eds.), Social Capital: A Multifaceted
Perspective, The World Bank: Washington.
Deci, E., and R. Ryan (1985), Intrinsic Motivation and Self-Determination in
Human Behavior, Plenum Press, New York.
Deci, E., R. Koestner, and R. Ryan (1999), A Meta-Analytic Review of Experiments Examining the Effects of Extrinsic Rewards on Intrinsic Motivation, Psychological Bulletin 125 (6), 627-668.
Diamond, P. (1982), Aggregate-Demand Management in Search Equilibrium,
Journal of Political Economy 90 (5), 881-894.
Diamond, P. (1984), Money in Search Equilibrium, Econometrica 52 (1), 1-20.
References
192
Dibb, S., L. Sinhin, W. Pride and O. Ferrell (1997), Marketing: Concepts
and Strategies, Houghton Mifflin, Boston (3rd European ed.).
Dickens, W. (1986), Crime and punishment again: the economic approach with
a psychological twist, Journal of Public Economics 30, 97-107.
Diecidue, E., and J. van de Ven (2003), Aspiration Levels in Decision Making, Mimeo INSEAD/Tilburg University.
Diecidue, E., and P. Wakker (2001), On the Intuition of Rank-Dependent
Utility, The Journal of Risk and Uncertainty 23 (3), 281-298.
Douglas, M. and B. Isherwoord (1978), The world of goods, Allen lane (ed.
1979).
Drucker, P. and R. Heizer (1967), To make my name good, University of
California Press.
Economist, The (2003), Regulating the diamond trade: A crook’s best friend,
January 4th.
Etzioni, A. (1988), The Moral Dimension: toward a new economics, the Free
Press.
Falk, A., E. Fehr, and U. Fischbacher (1999), On the Nature of Fair Behavior, Working Paper no. 17 University of Zürich.
Fehr, E. and S. Gächter (1999), Reciprocity and Economics, Implications of
Reciprocal behaviour for labour markets, Mimeo University of Zurich.
Fehr, E., and S. Gächter (2000), Fairness and Retaliation: The Economics of
Reciprocity, Journal of Economic Perspectives, 14, 159—181.
Fehr, E., and K. Schmidt (1999), A Theory of Fairness, Competition, and
Cooperation, Quarterly Journal of Economics, 114, 817—868.
Festinger, L. (1957), A theory of cognitive dissonance, Row, Peterson and Company.
Festinger, L., and J. Carlsmith (1959), Cognitive consequences of forced compliance, Journal of Abnormal and Social Psychology 58, 203-210.
193
References
Forge, A. (1972), The Golden Fleece, Man 15 (4), 527-540.
Frank, R. (1985), Choosing the Right Pond: Human Behavior and the Quest for
Status, Oxford University Press: Oxford.
Frank, R. (1988), Passions within Reason: the strategic role of emotions, W.W.
Norton & Company, New York.
Frank, R. (1997), The Frame of Reference as a Public Good, The Economic
Journal 107 (445), 1832-1847
Freedman, J. (1963), Long-term behavioral effects of cognitive dissonance, Journal of Experimental Social Psychology 1, 145-155.
Frey, B. (1997a), A Constitution for Knaves Crowds Out Civic Virtues, The
Economic Journal, 107, 1043—1053.
Frey, B. (1997b), Not Just for the Money: An Economic Theory of Personal
Motivation, Edward Elgar: Brookfield.
Frey, B., and R. Jegen (2002), Motivation Crowding Theory: A Survey of Empricial Evidence, Journal of Economic Surveys, forthcoming.
Friedman, M., and L. Savage (1948), The Utility Analysis of Choices Involving Risk, Journal of Political Economy 56 (4), 279-304.
Fudenberg, D., and J. Tirole (1991), Game Theory, MIT Press, Cambridge
MA..
Fukuyama, F. (1999), The great disruption; Human nature and the reconstruction of social order, the Free Press, New York.
Gächter, S., and E. Fehr (1999), Collective Action as a Social Exchange, Journal of Economic Behavior & Organization 39, 341—369.
Gächter, S., and A. Falk (1999), Reputation or Reciprocity,Working Paper
19, University of Zurich, Zurich.
Galor, O., and J. Zeira (1993), Income Distribution and Macroeconomics, Review of Economic Studies, 60, 35—52.
References
194
Glazer, A., and K. Konrad (1996), A Signaling Explanation for Charity, American Economic Review, 86, 1019—1028.
Gneezy, U., and A. Rustichini (2000), A Fine is a Price, The Journal of Legal Studies, 29, 1—18.
Goeree, J., C. Holt, and S. Laury (2002), Private Costs and Public Benefits: Unraveling the Effects of Altruism and Noisy Behavior, Journal of
Public Economics, 83, 255-276.
Goldman, S. (1978), Comment: gift equilibria and Pareto optimality, Journal
of Economic Theory 18, 368-370.
Grant, S., A. Kajii, and B. Polak (1998), Intrinsic Preference for Information, Journal of Economic Theory 83, 233-259.
Grant, S., A. Kajii, and B. Polak (2000), Preference for Information and Dynamic Consistency, Theory and Decision 48, 263-286.
Gregory, C. (1989), Gifts,109—118 in: J. Eatwell et al. (eds.), The New Palgrave
Social Economics, Macmillan: London.
Güth, W. (1995), On Ultimatum Bargaining Experiments: A personal review,
Journal of Economic Behaviour and Organization 27, 329-344.
Güth, W., and H. Kliemt (1994), Competition or Co-Operation: On the Evolutionary Economics of Trust, Exploitation and Moral Attitudes, Metroeconomica, 45, 155—187.
Güth, W., and H. Kliemt (1998), The Indirect Evolutionary Approach, Rationality and Society, 10, 377—399.
Güth, W. and E. van Damme (1998), Information, Strategic Behavior, and
Fairness in ultimatum bargaining: an experimental study, Journal of Mathematical Psychology 42, 227-247.
Harbaugh, T. (1998), The Prestige Motive for Making Charitable Transfers,
American Economic Review, 88, 277—282.
Harsanyi, J. (1969), Rational-Choice Models of Political Behavior vs. Functionalist and Conformist Theories, World Politics, 22, 513—538.
195
References
Henry, O. (1906) The Gift of the Magi and Other Short Stories, Dover Thrift,
New York (ed. 1992).
Hirsch, F. (1976), Social Limits to Growth, Harvard University Press: Cambridge, MA.
Holländer, H. (1990), A Social Exchange Approach to Voluntary Cooperation,
American Economic Review, 80, 1157—1167.
Homans, G. (1958),Social Behavior as Exchange, American Journal of Sociology, 63, 597—606.
Homans, G. (1961), Social Behaviour: Its Elementary Forms, Routlegde & Kegan
Paul: London.
Hume, D. (1886), A Treatise of Human Nature, Book II, in T. Green and T.
Grose (eds.), Hume: The Philosophical Works, Scienta Verlag Aalen.
Iannaccone, L. (1992), Sacrifice and Stigma: reducing free-riding in cults, communes, and other collectives, Journal of Political Economy 100 (2), 271-291.
James, J., and E. Gutkind (1985), Attitude change revisited: Cognitive dissonance theory and development policy, World Development 13 (10/11),
1139-1149.
Jones, E., and R. Kohler (1959), The effects of plausibility on the learning of
controversial statements, Journal of Abnormal and Social Psychology 57,
315-320.
Kahneman, D. (1994), New challenges to the rationality assumption, Journal
of Institutional and Theoretical Economics 150, 18-36.
Kahneman, D. (2000), Experienced Utility and Objective Happiness: A momentbased approach, in: Kahneman, D. and A. Tversky (eds.), Choices, Values,
and Frames, Cambridge University Press, New York.
Kahneman, D., and A. Tversky (1979), Prospect theory: an analysis of decision under risk, Econometrica 47 (2), 263-291.
References
196
Kahneman, D., P. Wakker, and R. Sarin (1997) Back to Bentham? Explorations of Experienced Utility, Quarterly Journal of Economics 112 (2),
375-405.
Kenrick, D., S. Neuberg, and R. Cialdini (1999), Social Psychology, Allyn
and Bacon: Boston.
Keyzer, M. (2002), Labeling and the realization of cultural values, De Economist 150 (4), 487-511.
Khalil, E. (1997), Etzioni versus Becker: Do moral sentiments differ from ordinary tastes?, De Economist 145 (4), 491-520.
Kirchhoff, S. (2000), Green Business and Blue Angels, Environmental and Resource Economics 15, 403-420.
Kohn, A. (1993), Punished by Rewards, Houghton-Mifflin, Boston.
Kranich, L. (1994), Gift Equilibria and Pareto Optimality Reconsidered, Journal of Economic Theory, 64, 298—300.
Kranton, R. (1996), Reciprocal Exchange: A Self-Sustaining System, American
Economic Review, 86, 830—851.
Kranton, R. (1996b), The formation of cooperative relationships, Journal of
Law, Economics, & Organization, 214-233.
Kropotkin, P. (1904), Mutual Aid; A factor of evolution, William Heinemann,
London.
Laibson, D. (1997), Golden Eggs and Hyberbolic Discounting, Quarterly Journal of Economics, 443-477.
Lancaster, K. (1966), A New Approach to Consumer Theory, Journal of Political Economy, 74, 132—157.
Leibenstein, H. (1950), Bandwagon, snob, and Veblen effects in the theory of
consumer’s demand, Quarterly Journal of Economics 64 (2), 183-207.
197
References
Lepper, M., D. Greene, and R. Nisbett (1973), Undermining Children’s Interest with Ectrinsic Reward: A test of the "overjustification" hypothesis,
Journal of Personality and social Psychology 28 (1), 129-173.
Loewenstein, G., and D. Adler (1995), A bias in the prediction of tastes,
The Economic Journal 105, 929-937.
Loomes, G. (1998), Probabilities vs Money: A test of some fundamental assumptions about rational decision making, The Economic Journal, 108,
477-489.
Lopes, L. (1984), Risk and Distributional Inequality, Journal of Experimental
Psychology: Human Perception and Performance 10 (4), 465-485.
Magat, W. and W. Viscusi (1992), Informational Approaches to Regulation,
MIT Press, Cambridge, Mass.
Mandeville, B. (1714) The fable of the bees: or, private vices, public benefits,
Oxford University Press, London (ed. 1957).
Mankiw, G. (1998), Principles of Economics, Dryden Press.
Mas-Colell, A., M. Whinston, and J. Green (1995), Microeconomic Theory,
Oxford University Press, New York.
Mason, C. (2002), On the Economics of Eco-Labeling, Mimeo University of
Wyoming.
Mauss, M. (1925/1980), The Gift, translated from Essai sur le don, Routledge
and Kegan Paul: London.
Milgrom, P. and J. Roberts (1986), Price and Advertising Signals of Product
Quality, Journal of Political Economy 94 (4), 796-821.
Mullanaithan, S., and R. Thaler (2000), Behavioral Economics, NBER Working Paper 7948.
Nagel, R. (1995), Unraveling in Guesssing Games, An experimental Study, American Economic Review 85(5), 1313-1326.
References
198
Ng, Y. (1997), A Case for Happiness, Cardinalism, and Interpersonal Comparability, The Economic Journal 107 (445), 1848-1858
Payne, J., D. Laughhunn, and R. Crum (1980), Translation of gambles and
aspiration level effects in risky choice behavior, Management Science 26
(10), 1039-1060.
Payne, J., D. Laughhunn, and R. Crum (1981), Further tests of aspiration
level effects in risky choice behavior, Management Science 27 (8), 953-958.
Polanyi, K. (1957), The Great Transformation, Beacon Press: Boston.
Prendergast, C., and L. Stole (2001), The non-monetary nature of gifts, European Economic Review 45 (10), 793-1810.
Pruitt, D. (1968), Reciprocity and Credit Building in a Laboratory Dyad, Journal of Personality and Social Psychology, 8, 143—147.
Rabin, M. (1993), Incorporating fairness into game theory and economics, American Economic Review 83 (5), 1281-1301.
Rabin, M. (1994), Cognitive dissonance and social change, Journal of Economic
Behavior and Organization 23, 177-194
Rabin, M. (1995), Moral preferences, moral constraints, and self-serving biases,
Berkeley Department of Economics Working Paper No 95-241.
Rabin, M. (1998), Psychology and Economics, Journal of Economic Literature
XXXVI, 11-46.
Rabin, M. (2000), Diminishing Marginal Utility of Wealth Cannot Explain Risk
Aversion, in Kahneman, D., and A. Tversky (eds.), Choices, Values, and
Frames, Cambridge University Press, NeW York.
Rabin, M. (2002), A Perspective on Psychology and Ecpnomics, European Economic Review 46, 657-685.
Robben, H., and T. Verhallen (1994), Behavioral Costs as Determinants of
Cost Perception and Preference Formation for Gifts to Receive and Gifts
to Give, Journal of Economic Psychology, 15, 333—350.
199
References
Roth, A (1995), Introduction to Experimental Economics, in:Kagel, J. and A.
Roth (eds.), Handbook of Experimental Economics, Princeton University
Press, Princeton.
Rubinstein, A. (2001), A theorist’s view of experiments, European Economic
Review 45, 651-218.
Ruffle, B. (1999), Gift-giving with emotions, Journal of Economic Behavior and
Organization 39, 399-420.
Sapolsky, R. (1998), Why zebras don’t get ulcers, Freeman, New York.
Satow, K. (1975), Social Approval and Helping, Journal of Experimental Social
Psychology, 11, 501—509.
Schwartz, B. (1967), The Social Psychology of the Gift, The American Journal
of Sociology, 73, 1—11.
Slovic, P. (1995), The construction of preference, American Psychologist 50 (5),
364-371.
Smith, V. (1962), An Experimental Study of Competitive Behavior, Journal of
Political Economy 70 (2), 111-137.
Solnick, S., and D. Hemenway (1996), The Deadweight Loss of Christmas:
Comment, American Economic Review, 86, 1299—1305.
Souvorov, A. (2003), Addiction to Rewards, Mimeo GREMAQ, Toulouse.
Spence, M. (1973), Job market signaling, Quarterly Journal of Economics 87,
355-374.
Starmer, C. (1998), Developments in Non-Expected Utility Theory: The Hunt
for a Descriptive Theory of Choice under Risk, Journal of Economic Literature XXXVIII, 332-382.
Sugden, R. (1989), Spontaneous Order, Journal of Economic Perspectives, 3,
85—97.
Tirole, J. (1988), The Theory of Industrial Organisation, MIT Press, Cambridge, Mass.
References
200
Tirole, J. (2002), Rational irrationality: Some economics of self-management,
European Economic Review 46, 633-655.
Thaler, R. (1980), Toward a positive Theory of Consumer Choice, Journal of
Economic Behavior and Organization 1, 39-60.
Thaler, R. (1999), Mental Accounting Matters, Journal of Behavioral Decision
Making 12, 183-206.
Thaler, R. (2000), From Homo Economicus to Homo Sapiens, Journal of Economic Perspectives 14 (1), 133-141.
Titmuss, R. (1970), The Gift Relationship, George Allen & Unwin Ltd.: London.
Trivers, R. (1971), The evolution of reciprocal altruism, Quarterly Review of
Biology 46, 35-57.
Tversky, A., and D. Kahneman (1992), Advances in prospect theory, cumulative representation of uncertainty, Journal of Risk and Uncertainty 5,
297-323.
Van Damme, E. (1989), Stable equilibria and forward induction, Journal of
Economic Theory 48 (2), 476-496.
VandeVelde, A. (2000), Towards a Conceptual Map of Gift Practices, in Gifts
and Interests, A. Vandevelde (ed.), Peeters.
Van Kempen, L. (2003), Fooling the eye of the Beholder: Deceptive Status Signalling among the Poor in Developing Countries, Journal of International
Development, 15, 157-177.
Varian, H. (1992), Microeconomic Analysis, W.W. Norton & Company, New
York (3rd ed.).
Veblen, T. (1899), The theory of the leisure class : an economic study of institutions, The New American Library, New York (ed. 1953).
Wakker, P. (1988), Nonexpected Utility as Aversion of Information, Journal of
Behavioral Decision Making 1, 169-175.
201
References
Waldfogel, J. (1993), The Deadweight Loss of Christmas, American Economic
Review, 83, 1329—1336.
Waldfogel, J. (1996), The Deadweight Loss of Christmas: Reply, American Economic Review, 86, 1306—1308.
Webley, P., and R. Wilson (1989), Social Relationships and the Unacceptability of Money as a Gift, Journal of Social Psychology, 129, 85—91.
Williamsion, O. (1985), The Economic Institutions of Capitalism, Free Press,
New York.
Wright, R. (1994), The Moral Animal, Brown & Company: London.
Yellen, J. (1990), The Transformation of the Kalahari !Kung Tribe, Scientific
American, 262, April, 96—105.
Zwick, R., I. Erev, and D. Budescu (1999), The Psychological and Economical Perspectives on Human Decisions in Social and Interactive Contexts,
in Budescu, D., I. Erev, and R. Zwick (eds.), Games and Human Behavior:
Essays in honor of Amnon Rapoport, Lawrence Erlbaum Associates, New
Jersey.
Center for Economic Research, Tilburg University, The Netherlands
Dissertation Series
No. Author
Title
Published
1 P.J.J. Herings
Static and Dynamic Aspects of General
Disequilibrium Theory; ISBN 90 5668 001 3
June
1995
21 Erwin van der
Krabben
Urban Dynamics: A Real Estate Perspective An institutional analysis of the production of
the built environment; ISBN 90 5170 390 2
August
1995
3 Arjan Lejour
Integrating or Desintegrating Welfare States? a qualitative study to the consequences of
economic integration on social insurance;
ISBN 90 5668 003 X
September
1995
4 Bas J.M. Werker
Statistical Methods in Financial Econometrics;
ISBN 90 5668 002 1
September
1995
5 Rudy Douven
Policy Coordination and Convergence in the
EU; ISBN 90 5668 004 8
September
1995
6 Arie J.T.M. Weeren
Coordination in Hierarchical Control;
ISBN 90 5668 006 4
September
1995
7 Herbert Hamers
Sequencing and Delivery Situations: a Game
Theoretic Approach; ISBN 90 5668 005 6
September
1995
8 Annemarie ter Veer
Strategic Decision Making in Politics;
ISBN 90 5668 007 2
October
1995
9 Zaifu Yang
Simplicial Fixed Point Algorithms and
Applications; ISBN 90 5668 008 0
January
1996
10 William Verkooijen
Neural Networks in Economic Modelling - An
Empirical Study; ISBN 90 5668 010 2
February
1996
11 Henny Romijn
Acquisition of Technological Capability in
Small Firms in Developing Countries;
ISBN 90 5668 009 9
March
1996
12 W.B. van den Hout
The Power-Series Algorithm - A Numerical
Approach to Markov Processes;
ISBN 90 5668 011 0
March
1996
13 Paul W.J. de Bijl
Essays in Industrial Organization and
Management Strategy; ISBN 90 5668 012 9
April
1996
No. Author
14 Martijn van de Ven
Title
Published
Intergenerational Redistribution in
Representative Democracies;
ISBN 90 5668 013 7
May
1996
15 Eline van der Heijden Altruism, Fairness and Public Pensions: An
Investigation of Survey and Experimental
Data; ISBN 90 5668 014 5
May
1996
16 H.M. Webers
Competition in Spatial Location Models;
ISBN 90 5668 015 3
June
1996
17 Jan Bouckaert
Essays in Competition with Product
Differentiation and Bargaining in Markets;
ISBN 90 5668 016 1
June
1996
18 Zafar Iqbal
Three-Gap Analysis of Structural Adjustment
in Pakistan; ISBN 90 5668 017 X
September
1996
19 Jimmy Miller
A Treatise on Labour: A Matching-Model
Analysis of Labour-Market Programmes;
ISBN 90 5668 018 8
September
1996
20 Edwin van Dam
Graphs with Few Eigenvalues - An interplay
between combinatorics and algebra;
ISBN 90 5668 019 6
October
1996
21 Henk Oosterhout
Takeover Barriers: the good, the bad, and the
ugly; ISBN 90 5668 020 X
November
1996
22 Jan Lemmen
Financial Integration in the European Union:
Measurement and Determination;
ISBN 90 5668 021 8
December
1996
23 Chris van Raalte
Market Formation and Market Selection;
ISBN 90 5668 022 6
December
1996
24 Bas van Aarle
Essays on Monetary and Fiscal Policy
Interaction: Applications to EMU and Eastern
Europe; ISBN 90 5668 023 4
December
1996
25 Francis Y. Kumah
Common Stochastic Trends and Policy Shocks
in the Open Economy: Empirical Essays in
International Finance and Monetary Policy;
ISBN 90 5668 024 2
May
1997
26 Erik Canton
Economic Growth and Business Cycles;
ISBN 90 5668 025 0
September
1997
No. Author
Title
Published
27 Jeroen
Hoppenbrouwers
Conceptual Modeling and the Lexicon;
ISBN 90 5668 027 7
October
1997
28 Paul Smit
Numerical Analysis of Eigenvalue Algorithms
Based on Subspace Iterations;
ISBN 90 5668 026 9
October
1997
29 Uri Gneezy
Essays in Behavioral Economics;
ISBN 90 5668 028 5
October
1997
30 Erwin Charlier
Limited Dependent Variable Models for Panel
Data; ISBN 90 5668 029 3
November
1997
31 Rob Euwals
Empirical Studies on Individual Labour
Market Behaviour; ISBN 90 5668 030 7
December
1997
32 Anurag N. Banerjee
The Sensitivity of Estimates, Inferences, and
Forecasts of Linear Models;
ISBN 90 5668 031 5
December
1997
33 Frans A. de Roon
Essays on Testing for Spanning and on
Modeling Futures Risk Premia;
ISBN 90 5668 032 3
December
1997
34 Xiangzhu Han
Product Differentiation, Collusion and
Standardization; ISBN 90 5668 033 1
January
1998
35 Marcel Das
On Income Expectations and Other Subjective
Data: A Micro-Econometric Analysis;
ISBN 90 5668 034 X
January
1998
36 Jeroen Suijs
Cooperative Decision Making in a Stochastic
Environment; ISBN 90 5668 035 8
March
1998
37 Talitha Feenstra
Environmental Policy Instruments and
International Rivalry: A Dynamic Analysis;
ISBN 90 5668 036 6
May
1998
38 Jan Bouwens
The Use of Management Accounting Systems
in Functionally Differentiated Organizations;
ISBN 90 5668 037 4
June
1998
39 Stefan Hochguertel
Households’ Portfolio Choices;
ISBN 90 5668 038 2
June
1998
40 Henk van Houtum
The Development of Cross-Border Economic
Relations; ISBN 90 5668 039 0
July
1998
41 Jorg Jansen
Service and Inventory Models subject to a
Delay-Limit; ISBN 90 5668 040 4
September
1998
No. Author
Title
Published
42 F.B.S.L.P. Janssen
Inventory Management Systems: control and
information issues; ISBN 90 5668 041 2
September
1998
43 Henri L.F. de Groot
Economic Growth, Sectoral Structure and
Unemployment; ISBN 90 5668 042 0
October
1998
44 Jenke R. ter Horst
Longitudinal Analysis of Mutual Fund
Performance; ISBN 90 5668 043 9
November
1998
45 Marco Hoeberichts
The Design of Monetary Institutions;
ISBN 90 5668 044 7
December
1998
46 Adriaan Kalwij
Household Consumption, Female Employment
and Fertility Decisions: A microeconometric
analysis; ISBN 90 5668 045 5
February
1999
47 Ursula Glunk
Realizing High Performance on Multiple
Stakeholder Domains: A Resource-based
Analysis of Professional Service Firms in the
Netherlands and Germany;
ISBN 90 5668 046 3
April
1999
48 Freek Vermeulen
Shifting Ground: Studies on the Intersection of February
Organizational Expansion, Internationalization, 1999
and Learning;
ISBN 90 5668 047 1
49 Haoran Pan
Competitive Pressures on Income Distribution
in China; ISBN 90 5668 048 X
March
1999
50 Laurence van Lent
Incomplete Contracting Theory in Empirical
Accounting Research; ISBN 90 5668 049 8
April
1999
51 Rob Aalbers
On the Implications of Thresholds for
Economic Science and Environmental Policy;
ISBN 90 5668 050 1
May
1999
52 Abe de Jong
An Empirical Analysis of Capital Structure
Decisions in Dutch Firms;
ISBN 90 5668 051 X
June
1999
53 Trea Aldershof
Female Labor Supply and Housing Decisions;
ISBN 90 5668 052 8
June
1999
54 Jan Fidrmuc
The Political Economy of Reforms in Central
and Eastern Europe; ISBN 90 5668 053 6
June
1999
55 Michael Kosfeld
Individual Decision-Making and Social
Interaction; ISBN 90 5668 054 4
June
1999
No. Author
Title
Published
56 Aldo de Moor
Empowering Communities - A method for the
legitimate user-driven specification of network
information systems; ISBN 90 5668 055 2
October
1999
57 Yohane A. Khamfula
Essays on Exchange Rate Policy in Developing September
Countries; ISBN 90 5668 056 0
1999
58 Maurice Koster
Cost Sharing in Production Situations and
Network Exploitation; ISBN 90 5668 057 9
59 Miguel Rosellón
Cifuentes
Essays on Financial Policy, Liquidation Values December
and Product Markets;
1999
ISBN 90 5668 058 7
60 Sharon Schalk
Equilibrium Theory: A salient approach;
ISBN 90 5668 059 5
December
1999
61 Mark Voorneveld
Potential Games and Interactive Decisions
with Multiple Criteria; ISBN 90 5668 060 9
December
1999
62 Edward Droste
Adaptive Behavior in Economic and Social
Environments; ISBN 90 5668 061 7
December
1999
63 Jos Jansen
Essays on Incentives in Regulation and
Innovation; ISBN 90 5668 062 5
January
2000
64 Franc J.G.M.
Klaassen
Exchange Rates and their Effects on
International Trade; ISBN 90 5668 063 3
January
2000
65 Radislav Semenov
Cross-country Differences in Economic
Governance: Culture as a major explanatory
factor; ISBN 90 5668 065 X
February
2000
66 Alexandre
Possajennikov
Learning and Evolution in Games and
Oligopoly Models; ISBN 90 5668 064 1
March
2000
67 Eric Gaury
Designing Pull Production Control Systems:
Customization and Robustness;
ISBN 90 5668 066 8
March
2000
68 Marco Slikker
Decision Making and Cooperation
Restrictions; ISBN 90 5668 067 6
April
2000
69 Ruud Brekelmans
Stochastic Models in Risk Theory and
Management Accounting;
ISBN 90 5668 068 4
June
2000
70 Flip Klijn
A Game Theoretic Approach to Assignment
Problems; ISBN 90 5668 069 2
June
2000
December
1999
No. Author
Title
Published
71 Maria Montero
Endogenous Coalition Formation and
Bargaining; ISBN 90 5668 070 6
June
2000
72 Bas Donkers
Subjective Information in Economic Decision
Making; ISBN 90 5668 071 4
June
2000
73 Richard Nahuis
Knowledge and Economic Growth;
ISBN 90 5668 072 2
October
2000
74 Kuno Huisman
Technology Investment: A Game Theoretic
Real Options Approach; ISBN 90 5668 073 0
October
2000
75 Gijsbert
van Lomwel
Essays on Labour Economics;
ISBN 90 5668 074 9
November
2000
76 Tina Girndt
Cultural Diversity and Work-Group
Performance: Detecting the Rules;
ISBN 90 5668 075 7
December
2000
77 Teye A. Marra
The Influence of Proprietary Disclosure Costs
on the Decision to Go Public: Theory and
Evidence; ISBN 90 5668 076 5
January
2001
78 Xiaodong Gong
Empirical Studies on the Labour Market and
on Consumer Demand; ISBN 90 5668 078 1
February
2001
79 Kanat Çamlibel
Complementarity Methods in the Analysis of
Piecewise Linear Dynamical Systems;
ISBN 90 5668 079 X
May
2001
80 Joost Driessen
Empirical Studies on the Pricing of Bonds and
Interest Rate Derivatives; ISBN 90 5668 080 3
June
2001
81 Bram van den Broek
Uncertainty in Differential Games;
ISBN 90 5668 081 1
June
2001
82 Ricardo Recht
Organisational Culture and Privatisation: A
June
Case Study of the Argentinean Railway Sector; 2001
ISBN 90 5668 082 X
83 Willem Verhagen
Inflation Targeting and Interest Rate Policy;
ISBN 90 5668 083 8
June
2001
84 Vincent Verouden
Essays in Antitrust Economics;
ISBN: 90 5668 086 2
June
2001
85
Rosalia
Vazquez Alvarez
A Nonparametric Approach to the Sample
Selection Problem in Survey Data;
ISBN 90 5668 087 0
June
2001
86
Theo Leers
Public Pensions and Population Ageing: An
Economic Analysis of Fertility, Migration, and
Social-Security Policy; ISBN 90 5668 084 6
June
2001
No.
Author
Title
Published
87
Bernard Conlon
Consumer Rationality in Choice;
ISBN 90 5668 085 4
June
2001
88
Judith Timmer
Cooperative behaviour, uncertainty and
operations research;
ISBN 90 5668 088 9
September
2001
89
Enrico Diecidue
Nonexpected Utility and Coherence;
ISBN 90 5668 089 7
November
2001
90 Alexander
Konovalov
Essays in General Equilibrium Theory;
ISBN 90 5668 091 9
November
2001
91 Stefan Stremersch
Essays on Marketing Strategy in TechnologyIntensive Markets; ISBN 90 5668 090 0
December
2001
92 Rian Drogendijk
Expansion Patterns of Dutch Firms in Central
and Eastern Europe: Learning to
Internationalize; ISBN 90 5668 092 7
December
2001
93 Han Donker
Takeovers. An Empirical Study of Ownership
Structure and the Wealth to Shareholders in
Dutch Takeovers; ISBN 90 5668 093 5
December
2001
94 Erica van Herpen
Perceptions and Evaluations of Assortment
Variety; ISBN 90 5668 094 3
December
2001
95 Jana Vyrastekova
Cheap Talk and Spiteful Preferences in
Ultimatum Games: Experiments and
Evolutionary Rationale; ISBN 90 5668 095 1
March
2002
96 Tao Lin
Statistics of Extremes in the Space of
Continuous Functions; ISBN 90 5668 096 X
Papers in Auction Theory;
ISBN 90 5668 097 8
May
2002
May
2002
June
2002
97 Sander Onderstal
98 Wolf Wagner
Risk Sharing under Incentive Constraints;
ISBN 90 5668 098 6
99
Integrating Modern Business Applications with
Objectified Legacy Systems;
ISBN 90 5668 099 4
June
2002
100 Jacco Wielhouwer
Optimal Tax Depreciation and its Effects on
Optimal Firm Investments;
ISBN 90 5668 100 1
June
2002
101 Peter Roosenboom
Corporate Governance Mechanisms in IPO
Firms; ISBN 90 5668 101 X
September
2002
Willem-Jan van den
Heuvel
No.
Author
Title
Published
102 Victoria Shestalova
Essays in Productivity and Efficiency;
ISBN 90 5668 103 6
September
2002
103 Paula van VeenDirks
Flexibility and Control: An Empirical Study
Relating Production Flexibility to the Design
of Performance Management Systems;
ISBN 90 5668 102 8
September
2002
104 Michal Matĕjka
Management Accounting in Organizational
Design: Three Essays; ISBN 90 5668 105 2
November
Market Liquidity around Earnings
Announcements; ISBN 90 5668 106 0
November
106 Nancy KampRoelands
Towards a framework for auditing
environmental reports; ISBN 90 5668 108 7
November
2002
107 Ana Ferreira
Statistics of extremes: estimation and
optimality; ISBN 90 5668 107 9
November
2002
108 Alexei Goriaev
On the behaviour of Mutual Fund Investors
and Managers; ISBN 90 5668 110 9
November
2002
109 Paul Ingenbleek
Money for Value Pricing from a ResourceAdvantage Perspective; ISBN 90 5668 109 5
December
2002
110 Frank Wijen
Stakeholder influence and organizational
learning in environmental management;
ISBN 90 5668 111 7
December
Labor market institutions in OECD countries:
Origins and consequences;
ISBN 90 5668 113 3
January
112 Federica Teppa
Estimating preference parameters in individual
financial decision making;
ISBN 90 5668 112 5
January
2003
113 Dmitry Danilov
The effects of pretesting in econometrics with
applications in finance;
ISBN 90 5668 104 4
February
2003
114 Jacco Thijssen
Investment under uncertainty, market evolution
and coalition spillovers in a game theoretic
perspective; ISBN 90 5668 114 1
May 2003
105 Maarten Pronk
111 Michele Belot
2002
2002
2002
2003
No.
1
Author
Title
Published
115 Laura Spierdijk
Empirical Studies of Market Microstructure;
ISBN 90 5668 115 X
June 2003
116 Nicole van Beeck
A New Decision Support Method for Local
Energy Planning in Developing Countries;
ISBN 90 5668 116 8
June 2003
117 Grzegorz Pawlina
Corporate Investment under Uncertainty and
Competition: A Real Options Approach;
ISBN 90 5668 118 4
June 2003
118 Jana P. Fidrmuc
The Impact of Blockholders on Information
Signalling, Productivity, and Managerial
Disciplining; ISBN 90 5668 117 6
June 2003
119
Culture and Economic Development in
Europe; ISBN 90 5668 121 4
September
2003
120 Bas van Groezen
The Wealth of Generations: Ageing, Pensions
and Welfare from a Macroeconomic
Perspective; ISBN 90 5668 120 6
October
2003
121 Pham Do Kim Hang
Essays in Game Theory and Natural Resource
Management; ISBN 90 5668 122 2
September
2003
122 Jeroen van de Ven
Psychological Sentiments and Economic
Behaviour; ISBN 90 5668 123 0
October
2003
Sjoerd Beugelsdijk
Copies can be ordered from Thela Thesis, Prinseneiland 305, 1013 LP Amsterdam, The Netherlands, phone: +
31 20 6255429; fax: +31 20 6203395; e-mail: [email protected]